Prirucnik PDF

Nuclear Medicine Physics
A Handbook for Teachers and Students
p(x)
f(x,y)
z
x
z
y
x
D.L. Bailey
J.L. Humm
A. Todd-Pokropek
A. van Aswegen
Technical Editors
Nuclear Medicine Physics:

A Handbook for Teachers and
Students
The following states are Members of the international atomic energy agency:
afGhaNisTaN
albaNia
alGeria
aNGola
arGeNTiNa
arMeNia
ausTralia
ausTria
aZerbaiJaN
bahaMas
bahraiN
baNGladesh
belarus
belGiuM
beliZe
beNiN
boliVia
bosNia aNd herZeGoViNa
boTsWaNa
braZil
bruNei darussalaM
bulGaria
burkiNa faso
buruNdi
caMbodia
caMerooN
caNada
ceNTral africaN
rePublic
chad
chile
chiNa
coloMbia
coNGo
cosTa rica
cTe diVoire
croaTia
cuba
cyPrus
cZech rePublic
deMocraTic rePublic
of The coNGo
deNMark
doMiNica
doMiNicaN rePublic
ecuador
eGyPT
el salVador
eriTrea
esToNia
eThioPia
fiJi
fiNlaNd
fraNce
GaboN
GeorGia
GerMaNy
GhaNa
Greece
GuaTeMala
haiTi
holy see
hoNduras
huNGary
icelaNd
iNdia
iNdoNesia
iraN, islaMic rePublic of
iraQ
irelaNd
israel
iTaly
JaMaica
JaPaN
JordaN
kaZakhsTaN
keNya
korea, rePublic of
kuWaiT
kyrGyZsTaN
lao PeoPles deMocraTic
rePublic
laTVia
lebaNoN
lesoTho
liberia
libya
liechTeNsTeiN
liThuaNia
luXeMbourG
MadaGascar
MalaWi
Malaysia
Mali
MalTa
Marshall islaNds
MauriTaNia, islaMic
rePublic of
MauriTius
MeXico
MoNaco
MoNGolia
MoNTeNeGro
Morocco
MoZaMbiQue
MyaNMar
NaMibia
NePal
NeTherlaNds
NeW ZealaNd
NicaraGua
NiGer
NiGeria
NorWay
oMaN
PakisTaN
Palau
PaNaMa
PaPua NeW GuiNea
ParaGuay
Peru
PhiliPPiNes
PolaNd
PorTuGal
QaTar
rePublic of MoldoVa
roMaNia
russiaN federaTioN
rWaNda
saN MariNo
saudi arabia
seNeGal
serbia
seychelles
sierra leoNe
siNGaPore
sloVakia
sloVeNia
souTh africa
sPaiN
sri laNka
sudaN
sWaZilaNd
sWedeN
sWiTZerlaNd
syriaN arab rePublic
TaJikisTaN
ThailaNd
The forMer yuGoslaV
rePublic of MacedoNia
ToGo
TriNidad aNd TobaGo
TuNisia
Turkey
uGaNda
ukraiNe
uNiTed arab eMiraTes
uNiTed kiNGdoM of
GreaT briTaiN aNd
NorTherN irelaNd
uNiTed rePublic
of TaNZaNia
uNiTed sTaTes of aMerica
uruGuay
uZbekisTaN
VeNeZuela, boliVariaN
rePublic of
VieT NaM
yeMeN
ZaMbia
ZiMbabWe
The agencys statute was approved on 23 october 1956 by the conference on the statute of the
iaea held at united Nations headquarters, New york; it entered into force on 29 July 1957. The
headquarters of the agency are situated in Vienna. its principal objective is to accelerate and enlarge the
contribution of atomic energy to peace, health and prosperity throughout the world.
Nuclear Medicine Physics:

A Handbook for Teachers and
Students
endorsed by:
AMERICAN ASSOCIATION OF PHYSICISTS IN MEDICINE,
ASIAOCEANIA FEDERATION OF ORGANIZATIONS
FOR MEDICAL PHYSICS,
AUSTRALASIAN COLLEGE OF PHYSICAL SCIENTISTS
AND ENGINEERS IN MEDICINE,
EUROPEAN FEDERATION OF ORGANISATIONS
FOR MEDICAL PHYSICS,
FEDERATION OF AFRICAN MEDICAL PHYSICS ORGANISATIONS,
WORLD FEDERATION OF NUCLEAR MEDICINE AND BIOLOGY
International atomic energy agency

Vienna, 2014
COPYRIGHT NOTICE
All IAEA scientific and technical publications are protected by the terms of
the Universal Copyright Convention as adopted in 1952 (Berne) and as revised
in 1972 (Paris). The copyright has since been extended by the World Intellectual
Property Organization (Geneva) to include electronic and virtual intellectual
property. Permission to use whole or parts of texts contained in IAEA publications
in printed or electronic form must be obtained and is usually subject to royalty
agreements. Proposals for non-commercial reproductions and translations are
welcomed and considered on a case-by-case basis. Enquiries should be addressed
to the IAEA Publishing Section at:
Marketing and Sales Unit, Publishing Section
International Atomic Energy Agency
Vienna International Centre
PO Box 100
1400 Vienna, Austria
fax: +43 1 2600 29302
tel.: +43 1 2600 22417
email: sales.publications@iaea.org
http://www.iaea.org/books
IAEA, 2014
Printed by the IAEA in Austria
December 2014
STI/PUB/1617
IAEA Library Cataloguing in Publication Data

Nuclear medicine physics : a handbook for students and teachers. Vienna :
International Atomic Energy Agency, 2014.
p. ; 24 cm.
STI/PUB/1617
ISBN 9789201438102
Includes bibliographical references. 1. Nuclear medicine Handbooks, manuals, etc.
2. Medical physics handbooks. 3. Medical physics. I. International Atomic Energy
Agency.
IAEAL1400880
FOREWORD
Nuclear medicine is the use of radionuclides in medicine for diagnosis,
staging of disease, therapy and monitoring the response of a disease process.
It is also a powerful translational tool in the basic sciences, such as biology, in
drug discovery and in pre-clinical medicine. Developments in nuclear medicine
are driven by advances in this multidisciplinary science that includes physics,
chemistry, computing, mathematics, pharmacology and biology.
This handbook comprehensively covers the physics of nuclear medicine.
It is intended for undergraduate and postgraduate students of medical physics.
It will also serve as a resource for interested readers from other disciplines, for
example, clinicians, radiochemists and medical technologists who would like to
familiarize themselves with the basic concepts and practice of nuclear medicine
physics.
The scope of the book is intentionally broad. Physics is a vital aspect of
nearly every area of nuclear medicine, including imaging instrumentation,
image processing and reconstruction, data analysis, radionuclide production,
radionuclide therapy, radiopharmacy, radiation protection and biology. The
authors were drawn from a variety of regions and were selected because of their
knowledge, teaching experience and scientific acumen.
This book was written to address an urgent need for a comprehensive,
contemporary text on the physics of nuclear medicine. It complements similar
texts in radiation oncology physics and diagnostic radiology physics that have
been published by the IAEA.
Endorsement of this handbook has been granted by the following
international professional bodies: the American Association of Physicists in
Medicine (AAPM), the AsiaOceania Federation of Organizations for Medical
Physics (AFOMP), the Australasian College of Physical Scientists and Engineers
in Medicine (ACPSEM), the European Federation of Organisations for Medical
Physics (EFOMP), the Federation of African Medical Physics Organisations
(FAMPO), and the World Federation of Nuclear Medicine and Biology
(WFNMB).
The following international experts are gratefully acknowledged
for making major contributions to this handbook as technical editors:
D.L. Bailey (Australia), J.L. Humm (United States of America), A. Todd-Pokropek
(United Kingdom) and A. van Aswegen (South Africa). The IAEA officers
responsible for this publication were S. Palm and G.L. Poli of the Division of
Human Health.
EDITORIAL NOTE
Although great care has been taken to maintain the accuracy of information contained
in this publication, neither the IAEA nor its Member States assume any responsibility for
consequences which may arise from its use.
The use of particular designations of countries or territories does not imply any
judgement by the publisher, the IAEA, as to the legal status of such countries or territories, of
their authorities and institutions or of the delimitation of their boundaries.
The mention of names of specific companies or products (whether or not indicated as
registered) does not imply any intention to infringe proprietary rights, nor should it be construed
as an endorsement or recommendation on the part of the IAEA.
The IAEA has no responsibility for the persistence or accuracy of URLs for external or
third party Internet web sites referred to in this book and does not guarantee that any content
on such web sites is, or will remain, accurate or appropriate.
Preface
Nuclear medicine is the study and utilization of radioactive compounds in
medicine to image and treat human disease. It relies on the tracer principle first
espoused by Georg Karl von Hevesy in the early 1920s. The tracer principle is
the study of the fate of compounds in vivo usingminute amounts of radioactive
tracers which do not elicit any pharmacological response by the body to the tracer.
Today, the same principle is used to study many aspects of physiology, such as
cellular metabolism, DNA (deoxyribonucleic acid) proliferation, blood flow in
organs, organ function, receptor expression and abnormal physiology, externally
using sensitive imaging devices. Larger amounts of radionuclides are also applied
to treat patients with radionuclide therapy, especially in disseminated diseases
such as advanced metastatic cancer, as this form of therapy has the ability to
target abnormal cells to treat the disease anywhere in the body.
Nuclear medicine relies on function. For this reason, it is referred to as
functional imaging. Rather than just imaging a portion of the body believed
to have some abnormality, as is done with X ray imaging in radiology, nuclear
medicine scans often depict the whole body distribution of the radioactive
compound often acquired as a sequence of images over time showing the
temporal course of the radiotracer in the body.
There are two main types of radiation of interest for imaging in nuclear
medicine: ray emission from excited nuclei, and annihilation (or coincidence)
radiation () arising after positron emission from proton-rich nuclei. Gamma
photons are detected with a gamma camera as either planar (2D) images or
tomographically in 3D using single photon emission computed tomography.
The annihilation photons from positron emission are detected using a positron
emission tomography (PET) camera. The most recent major development in this
field is the combination of gamma cameras or PET cameras with high resolution
structural imaging devices, either X ray computed tomography (CT) scanners
or, increasingly, magnetic resonance imaging (MRI) scanners, in a single image
device. The combined PET/CT (or PET/MRI) scanner represents one of the most
sophisticated and powerful ways to visualize normal and altered physiology in
the body.
It is in this complex environment that the medical physicist, along with
nuclear medicine physicians and technologists/radiographers, plays a significant
role in the multidisciplinary team needed for medical diagnosis. The physicist is
responsible for such areas as instrumentation performance, radiation dosimetry
for treatment of patients, radiation protection of staff and accuracy of the data
analysis. The physicist draws on training in radiation and nuclear science,
in addition to scientific rigour and attention to detail in experiments and
measurements, to join forces with the other members of the multidisciplinary
team in delivering optimal health care. Patients are frequently treated on the
basis of the result of the scans they receive and these, therefore, have to be of the
highest quality.
This handbook was conceived and written by physicists, and is intended
primarily for physicists, although interested readers from medical, paramedical
and other science and engineering backgrounds could find it useful. The level
of understanding of the material covered will be different depending on the
background of the reader. Readers are encouraged to visit the IAEA Human
Health web site (http://www-naweb.iaea.org/NAHU/index.html) to discover the
wealth of resources available.
The technical editors and authors, selected for their experience and in
recognition of their contributions to the field, were drawn from around the world
and, thus, this book represents a truly international collaboration. The technical
editors travelled to the IAEA headquarters in Vienna on four occasions over three
years to bring this project to fruition. We would like to thank all of the authors for
their important contribution.
D.L. Bailey, J.L. Humm
A. Todd-Pokropek, A. van Aswegen
Contents
CHAPTER 1. BASIC PHYSICS FOR NUCLEAR MEDICINE . . . . . . . . . 1
1.1.
1.2.
1.3.
1.4.
1.5.
INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.1. Fundamental physical constants . . . . . . . . . . . . . . . . . . . 1
1.1.2. Physical quantities and units. . . . . . . . . . . . . . . . . . . . . . 2
1.1.3. Classification of radiation. . . . . . . . . . . . . . . . . . . . . . . . 4
1.1.4. Classification of ionizing radiation. . . . . . . . . . . . . . . . . 4
1.1.5. Classification of indirectly ionizing photon radiation. . . 5
1.1.6. Characteristic X rays. . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.1.7. Bremsstrahlung . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.1.8. Gamma rays. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.1.9. Annihilation quanta . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.1.10. Radiation quantities and units. . . . . . . . . . . . . . . . . . . . . 7
BASIC DEFINITIONS FOR ATOMIC STRUCTURE. . . . . . . . 8
1.2.1. Rutherford model of the atom. . . . . . . . . . . . . . . . . . . . . 10
1.2.2. Bohr model of the hydrogen atom. . . . . . . . . . . . . . . . . . 10
BASIC DEFINITIONS FOR NUCLEAR STRUCTURE. . . . . . 10
1.3.1. Nuclear radius. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.3.2. Nuclear binding energy. . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.3.3. Nuclear fusion and fission. . . . . . . . . . . . . . . . . . . . . . . . 13
1.3.4. Two-particle collisions and nuclear reactions. . . . . . . . . 14
RADIOACTIVITY. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.4.1. Decay of radioactive parent into a stable or unstable
daughter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.4.2. Radioactive series decay . . . . . . . . . . . . . . . . . . . . . . . . 19
1.4.3. Equilibrium in parentdaughter activities. . . . . . . . . . . . 21
1.4.4. Production of radionuclides (nuclear activation) . . . . . . 22
1.4.5. Modes of radioactive decay . . . . . . . . . . . . . . . . . . . . . . 23
1.4.6. Alpha decay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
1.4.7. Beta minus decay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
1.4.8. Beta plus decay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
1.4.9. Electron capture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
1.4.10. Gamma decay and internal conversion . . . . . . . . . . . . . 27
1.4.11. Characteristic (fluorescence) X rays and
Augerelectrons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
ELECTRON INTERACTIONS WITH MATTER. . . . . . . . . . . . 29
1.5.1. Electronorbital interactions. . . . . . . . . . . . . . . . . . . . . . 29
1.5.2. Electronnucleus interactions. . . . . . . . . . . . . . . . . . . . . 29
1.6.
PHOTON INTERACTIONS WITH MATTER. . . . . . . . . . . . . . 30

1.6.1. Exponential absorption of photon beam in absorber. . . . 30
1.6.2. Characteristic absorber thicknesses . . . . . . . . . . . . . . . . 31
1.6.3. Attenuation coefficients . . . . . . . . . . . . . . . . . . . . . . . . . 34
1.6.4. Photon interactions on the microscopic scale. . . . . . . . . 35
1.6.5. Photoelectric effect. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
1.6.6. Rayleigh (coherent) scattering. . . . . . . . . . . . . . . . . . . . . 39
1.6.7. Compton effect (incoherent scattering). . . . . . . . . . . . . . 39
1.6.8. Pair production . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
1.6.9. Relative predominance of individual effects . . . . . . . . . 46
1.6.10. Macroscopic attenuation coefficients . . . . . . . . . . . . . . . 47
1.6.11. Effects following photon interactions with absorber
and summary of photon interactions . . . . . . . . . . . . . . . 48
CHAPTER 2. BASIC RADIOBIOLOGY. . . . . . . . . . . . . . . . . . . . . . . . . . . 49

2.1.
2.2.
2.3.
2.4.
2.5.
2.6.
INTRODUCTION. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
RADIATION EFFECTS AND TIMESCALES. . . . . . . . . . . . . . 49
BIOLOGICAL PROPERTIES OF IONIZING RADIATION . . 51
2.3.1. Types of ionizing radiation. . . . . . . . . . . . . . . . . . . . . . . 51
MOLECULAR EFFECTS OF RADIATION AND THEIR
MODIFIERS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
2.4.1. Role of oxygen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
2.4.2. Bystander effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
DNA DAMAGE AND REPAIR. . . . . . . . . . . . . . . . . . . . . . . . . . 55
2.5.1. DNA damage. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
2.5.2. DNA repair . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
CELLULAR EFFECTS OF RADIATION . . . . . . . . . . . . . . . . . 56
2.6.1. Concept of cell death. . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
2.6.2. Cell survival curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
2.6.3. Dose deposition characteristics: linear energy transfer. . 57
2.6.4. Determination of relative biological effectiveness. . . . . 58
2.6.5. The dose rate effect and the concept of repeat
treatments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
2.6.6. The basic linearquadratic model. . . . . . . . . . . . . . . . . . 63
2.6.7. Modification to the linearquadratic model for
radionuclide therapies. . . . . . . . . . . . . . . . . . . . . . . . . . . 64
2.6.8. Quantitative intercomparison of different treatment
types. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
2.6.9. Cellular recovery processes . . . . . . . . . . . . . . . . . . . . . . 65
2.6.10. Consequence of radionuclide heterogeneity. . . . . . . . . . 66
2.7.
2.8.
GROSS RADIATION EFFECTS ON TUMOURS AND

TISSUES/ORGANS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
2.7.1. Classification of radiation damage (early versus late) . . 66
2.7.2. Determinants of tumour response. . . . . . . . . . . . . . . . . . 67
2.7.3. The concept of therapeutic index in radiation
therapy and radionuclide therapy . . . . . . . . . . . . . . . . . . 68
2.7.4. Long term concerns: stochastic and deterministic
effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
SPECIAL RADIOBIOLOGICAL CONSIDERATIONS IN
TARGETED RADIONUCLIDE THERAPY. . . . . . . . . . . . . . . . 69
2.8.1. Radionuclide targeting . . . . . . . . . . . . . . . . . . . . . . . . . . 69
2.8.2. Whole body irradiation. . . . . . . . . . . . . . . . . . . . . . . . . . 69
2.8.3. Critical normal tissues for radiation and
radionuclide therapies. . . . . . . . . . . . . . . . . . . . . . . . . . . 70
2.8.4. Imaging the radiobiology of tumours . . . . . . . . . . . . . . . 71
2.8.5. Choice of radionuclide to maximize therapeutic index. . 71
CHAPTER 3. RADIATION PROTECTION. . . . . . . . . . . . . . . . . . . . . . . . . 73

3.1.
3.2.
3.3.
3.4.
INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
BASIC PRINCIPLES OF RADIATION PROTECTION . . . . . . 74
3.2.1. The International Commission on Radiological
Protection system of radiological protection. . . . . . . . . . 74
3.2.2. Safety standards. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
3.2.3. Radiation protection quantities and units . . . . . . . . . . . . 77
IMPLEMENTATION OF RADIATION PROTECTION IN
A NUCLEAR MEDICINE FACILITY . . . . . . . . . . . . . . . . . . . . 81
3.3.1. General aspects. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
3.3.2. Responsibilities. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
3.3.3. Radiation protection programme. . . . . . . . . . . . . . . . . . . 84
3.3.4. Radiation protection committee . . . . . . . . . . . . . . . . . . . 84
3.3.5. Education and training. . . . . . . . . . . . . . . . . . . . . . . . . . . 84
FACILITY DESIGN. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
3.4.1. Location and general layout . . . . . . . . . . . . . . . . . . . . . . 85
3.4.2. General building requirements . . . . . . . . . . . . . . . . . . . . 85
3.4.3. Source security and storage . . . . . . . . . . . . . . . . . . . . . . 86
3.4.4. Structural shielding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
3.4.5. Classification of workplaces . . . . . . . . . . . . . . . . . . . . . 87
3.4.6. Workplace monitoring . . . . . . . . . . . . . . . . . . . . . . . . . . 88
3.4.7. Radioactive waste . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
3.5.
3.6.
3.7.
3.8.
3.9.
OCCUPATIONAL EXPOSURE . . . . . . . . . . . . . . . . . . . . . . . . . 89
3.5.1. Sources of exposure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
3.5.2. Justification, optimization and dose limitation. . . . . . . . 91
3.5.3. Conditions for pregnant workers and young persons . . . 91
3.5.4. Protective clothing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
3.5.5. Safe working procedures. . . . . . . . . . . . . . . . . . . . . . . . . 92
3.5.6. Personal monitoring. . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
3.5.7. Monitoring of the workplace. . . . . . . . . . . . . . . . . . . . . . 95
3.5.8. Health surveillance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
3.5.9. Local rules and supervision. . . . . . . . . . . . . . . . . . . . . . . 96
PUBLIC EXPOSURE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
3.6.1. Justification, optimization and dose limitation . . . . . . . . 97
3.6.2. Design considerations. . . . . . . . . . . . . . . . . . . . . . . . . . . 97
3.6.3. Exposure from patients. . . . . . . . . . . . . . . . . . . . . . . . . . 98
3.6.4. Transport of sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
MEDICAL EXPOSURE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
3.7.1. Justification of medical exposure . . . . . . . . . . . . . . . . . . 99
3.7.2. Optimization of protection . . . . . . . . . . . . . . . . . . . . . . . 100
3.7.3. Helping in the care, support or comfort of patients. . . . . 107
3.7.4. Biomedical research . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
3.7.5. Local rules. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
POTENTIAL EXPOSURE. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
3.8.1. Safety assessment and accident prevention. . . . . . . . . . . 108
3.8.2. Emergency plans. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
3.8.3. Reporting and lessons learned. . . . . . . . . . . . . . . . . . . . . 111
QUALITY ASSURANCE. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
3.9.1. General considerations . . . . . . . . . . . . . . . . . . . . . . . . . . 112
3.9.2. Audit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
CHAPTER 4. RADIONUCLIDE PRODUCTION. . . . . . . . . . . . . . . . . . . . 117

4.1.
4.2.
THE ORIGINS OF DIFFERENT NUCLEI. . . . . . . . . . . . . . . . . 117

4.1.1. Induced radioactivity. . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
4.1.2. Nuclide chart and line of nuclear stability. . . . . . . . . . . . 120
4.1.3. Binding energy, Q-value, reaction threshold and
nuclear reaction formalism . . . . . . . . . . . . . . . . . . . . . . . 123
4.1.4. Types of nuclear reaction, reaction channels and
cross-section . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
REACTOR PRODUCTION. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
4.2.1. Principle of operation and neutron spectrum. . . . . . . . . 128
4.3.
4.4.
4.5.
4.2.2. Thermal and fast neutron reactions . . . . . . . . . . . . . . . . 128

4.2.3. Nuclear fission, fission products . . . . . . . . . . . . . . . . . . 131
ACCELERATOR PRODUCTION. . . . . . . . . . . . . . . . . . . . . . . . 132
4.3.1. Cyclotron, principle of operation,
negativeandpositiveions. . . . . . . . . . . . . . . . . . . . . . . . 134
4.3.2. Commercial production (low and high energy). . . . . . . . 136
4.3.3. In-house low energy production (PET). . . . . . . . . . . . . . 137
4.3.4. Targetry, optimizing the production regarding yield
and impurities, yield calculations . . . . . . . . . . . . . . . . . . 140
RADIONUCLIDE GENERATORS. . . . . . . . . . . . . . . . . . . . . . . 141
4.4.1. Principles of generators. . . . . . . . . . . . . . . . . . . . . . . . . . 142
RADIOCHEMISTRY OF IRRADIATED TARGETS. . . . . . . . . 143
4.5.1. Carrier-free, carrier-added systems. . . . . . . . . . . . . . . . . 144
4.5.2. Separation methods, solvent extraction, ion
exchange, thermal diffusion . . . . . . . . . . . . . . . . . . . . . . 145
4.5.3. Radiation protection considerations and hot-box
facilities. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
CHAPTER 5. STATISTICS FOR RADIATION MEASUREMENT . . . . . . 149

5.1.
5.2.
5.3.
5.4.
SOURCES OF ERROR IN NUCLEAR MEDICINE

MEASUREMENT. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
CHARACTERIZATION OF DATA. . . . . . . . . . . . . . . . . . . . . . . 153
5.2.1. Measures of central tendency and variability. . . . . . . . . 153
STATISTICAL MODELS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
5.3.1. Conditions when binomial, Poisson and normal
distributions are applicable. . . . . . . . . . . . . . . . . . . . . . . 158
5.3.2. Binomial distribution. . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
5.3.3. Poisson distribution. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
5.3.4. Normal distribution. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
ESTIMATION OF THE PRECISION OF A SINGLE
MEASUREMENT IN SAMPLE COUNTING AND
IMAGING . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
5.4.1. Assumption. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
5.4.2. The importance of the fractional F as an indicator
of the precision of a single measurement in sample
counting and imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
5.4.3. Caution on the use of the estimate of the precision
of a single measurement in sample counting and
imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
5.5.
5.6.
5.7.
PROPAGATION OF ERROR . . . . . . . . . . . . . . . . . . . . . . . . . . . 172

5.5.1. Sums and differences. . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
5.5.2. Multiplication and division by a constant. . . . . . . . . . . . 174
5.5.3. Products and ratios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
APPLICATIONS OF STATISTICAL ANALYSIS . . . . . . . . . . . 177
5.6.1. Multiple independent counts. . . . . . . . . . . . . . . . . . . . . . 177
5.6.2. Standard deviation and relative standard deviation
for counting rates. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
5.6.3. Effects of background counts . . . . . . . . . . . . . . . . . . . . . 179
5.6.4. Significance of differences between counting
measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
5.6.5. Minimum detectable counts, count rate and activity. . . . 184
5.6.6. Comparing counting systems . . . . . . . . . . . . . . . . . . . . . 187
5.6.7. Estimating required counting times. . . . . . . . . . . . . . . . . 188
5.6.8. Calculating uncertainties in the measurement of
plasma volume in patients. . . . . . . . . . . . . . . . . . . . . . . . 189
APPLICATION OF STATISTICAL ANALYSIS:
DETECTOR PERFORMANCE. . . . . . . . . . . . . . . . . . . . . . . . . . 191
5.7.1. Energy resolution of scintillation detectors. . . . . . . . . . . 191
5.7.2. Intervals between successive events. . . . . . . . . . . . . . . . 193
5.7.3. Paralysable dead time . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
CHAPTER 6. BASIC RADIATION DETECTORS . . . . . . . . . . . . . . . . . . . 196

6.1.
6.2.
6.3.
6.4.
INTRODUCTION. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
6.1.1. Radiation detectors complexity and relevance. . . . . . 196
6.1.2. Interaction mechanisms, signal formation and
detector type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
6.1.3. Counting, current, integrating mode. . . . . . . . . . . . . . . . 197
6.1.4. Detector requirements. . . . . . . . . . . . . . . . . . . . . . . . . . . 197
GAS FILLED DETECTORS. . . . . . . . . . . . . . . . . . . . . . . . . . . . 200
6.2.1. Basic principles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200
SEMICONDUCTOR DETECTORS. . . . . . . . . . . . . . . . . . . . . . 202
6.3.2. Semiconductor detectors. . . . . . . . . . . . . . . . . . . . . . . . . 204
SCINTILLATION DETECTORS AND STORAGE
PHOSPHORS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
6.4.2. Light sensors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
6.4.3. Scintillator materials. . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
CHAPTER 7. ELECTRONICS RELATED TO NUCLEAR

MEDICINE IMAGING DEVICES . . . . . . . . . . . . . . . . . . . . 214
7.1.
7.2.
7.3.
7.4.
7.5.
7.6.
7.7.
INTRODUCTION. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
PRIMARY RADIATION DETECTION PROCESSES. . . . . . . . 215
7.2.1. Scintillation counters. . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
7.2.2. Gas filled detection systems . . . . . . . . . . . . . . . . . . . . . . 216
7.2.3. Semiconductor detectors. . . . . . . . . . . . . . . . . . . . . . . . . 216
IMAGING DETECTORS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
7.3.1. The gamma camera. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
7.3.2. The positron camera . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
7.3.3. Multiwire proportional chamber based X ray and
ray imagers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
7.3.4. Semiconductor imagers. . . . . . . . . . . . . . . . . . . . . . . . . . 220
7.3.5. The autoradiography imager. . . . . . . . . . . . . . . . . . . . . . 221
SIGNAL AMPLIFICATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
7.4.1. Typical amplifier . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
7.4.2. Properties of amplifiers. . . . . . . . . . . . . . . . . . . . . . . . . . 224
SIGNAL PROCESSING . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
7.5.1. Analogue signal utilization. . . . . . . . . . . . . . . . . . . . . . . 226
7.5.2. Signal digitization. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
7.5.3. Production and use of timing information. . . . . . . . . . . . 228
OTHER ELECTRONICS REQUIRED BY IMAGING
SYSTEMS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
7.6.1. Power supplies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
7.6.2. Uninterruptible power supplies. . . . . . . . . . . . . . . . . . . . 231
7.6.3. Oscilloscopes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
SUMMARY. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232
CHAPTER 8. GENERIC PERFORMANCE MEASURES . . . . . . . . . . . . . 234

8.1.
8.2.
INTRINSIC AND EXTRINSIC MEASURES. . . . . . . . . . . . . . . 234

8.1.1. Generic nuclear medicine imagers . . . . . . . . . . . . . . . . . 234
8.1.2. Intrinsic performance. . . . . . . . . . . . . . . . . . . . . . . . . . . 236
8.1.3. Extrinsic performance. . . . . . . . . . . . . . . . . . . . . . . . . . . 236
ENERGY RESOLUTION. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
8.2.1. Energy spectrum. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
8.2.2. Intrinsic measurement energy resolution. . . . . . . . . . 238
8.2.3. Impact of energy resolution on extrinsic imager
performance. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
8.3.
8.4.
8.5.
8.6.
8.7.
SPATIAL RESOLUTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240

8.3.1. Spatial resolution blurring. . . . . . . . . . . . . . . . . . . . . . . . 240
8.3.2. General measures of spatial resolution. . . . . . . . . . . . . . 241
8.3.3. Intrinsic measurement spatial resolution . . . . . . . . . . 242
8.3.4. Extrinsic measurement spatial resolution. . . . . . . . . . 242
TEMPORAL RESOLUTION. . . . . . . . . . . . . . . . . . . . . . . . . . . 244
8.4.1. Intrinsic measurement temporal resolution . . . . . . . . 244
8.4.2. Dead time. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
8.4.3. Count rate performance measures. . . . . . . . . . . . . . . . . . 246
SENSITIVITY. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
8.5.1. Image noise and sensitivity. . . . . . . . . . . . . . . . . . . . . . . 247
8.5.2. Extrinsic measure sensitivity. . . . . . . . . . . . . . . . . . . 248
IMAGE QUALITY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
8.6.1. Image uniformity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
8.6.2. Resolution/noise trade-off. . . . . . . . . . . . . . . . . . . . . . . . 249
OTHER PERFORMANCE MEASURES. . . . . . . . . . . . . . . . . . 250
CHAPTER 9. PHYSICS IN THE RADIOPHARMACY. . . . . . . . . . . . . . . 251

9.1.
9.2.
9.3.
9.4.
9.5.
9.6.
THE MODERN RADIONUCLIDE CALIBRATOR. . . . . . . . . . 251

9.1.1. Construction of dose calibrators. . . . . . . . . . . . . . . . . . . 251
9.1.2. Calibration of dose calibrators. . . . . . . . . . . . . . . . . . . . 253
9.1.3. Uncertainty of activity measurements. . . . . . . . . . . . . . . 254
9.1.4. Measuring pure emitters. . . . . . . . . . . . . . . . . . . . . . . . 258
9.1.5. Problems arising from radionuclide contaminants . . . . . 259
DOSE CALIBRATOR ACCEPTANCE TESTING AND
QUALITY CONTROL. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 260
9.2.1. Acceptance tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 260
9.2.2. Quality control. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262
STANDARDS APPLYING TO DOSE CALIBRATORS . . . . . . 262
NATIONAL ACTIVITY INTERCOMPARISONS. . . . . . . . . . . 263
DISPENSING RADIOPHARMACEUTICALS FOR
INDIVIDUAL PATIENTS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 264
9.5.1. Adjusting the activity for differences in patient size
and weight. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 264
9.5.2. Paediatric dosage charts . . . . . . . . . . . . . . . . . . . . . . . . . 264
9.5.3. Diagnostic reference levels in nuclear medicine. . . . . . . 266
RADIATION SAFETY IN THE RADIOPHARMACY . . . . . . . 269
9.6.1. Surface contamination limits. . . . . . . . . . . . . . . . . . . . . . 269
9.6.2. Wipe tests and daily surveys. . . . . . . . . . . . . . . . . . . . . . 270
9.6.3. Monitoring of staff finger doses during dispensing . . . . 270
9.7.
PRODUCT CONTAINMENT ENCLOSURES. . . . . . . . . . . . . . 271

9.7.1. Fume cupboards. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271
9.7.2. Laminar flow cabinets. . . . . . . . . . . . . . . . . . . . . . . . . . . 272
9.7.3. Isolator cabinets. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273
9.8. SHIELDING FOR RADIONUCLIDES . . . . . . . . . . . . . . . . . . . 274
9.8.1. Shielding for , and positron emitters . . . . . . . . . . . . . 274
9.8.2. Transmission factors for lead and concrete. . . . . . . . . . . 278
9.9. DESIGNING A RADIOPHARMACY. . . . . . . . . . . . . . . . . . . . . 280
9.10. SECURITY OF THE RADIOPHARMACY. . . . . . . . . . . . . . . . 282
9.11. RECORD KEEPING. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283
9.11.1. Quality control records . . . . . . . . . . . . . . . . . . . . . . . . . . 283
9.11.2. Records of receipt of radioactive materials. . . . . . . . . . . 283
9.11.3. Records of radiopharmaceutical preparation and
dispensing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 284
9.11.4. Radioactive waste records. . . . . . . . . . . . . . . . . . . . . . . . 284
CHAPTER 10. NON-IMAGING DETECTORS AND COUNTERS. . . . . . . 287

10.1. INTRODUCTION. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 287
10.2. OPERATING PRINCIPLES OF RADIATION DETECTORS. . 287
10.2.1. Ionization detectors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 288
10.2.2. Scintillation detectors . . . . . . . . . . . . . . . . . . . . . . . . . . . 292
10.3. RADIATION DETECTOR PERFORMANCE. . . . . . . . . . . . . . 294
10.3.1. Sensitivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 294
10.3.2. Energy resolution. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295
10.3.3. Count rate performance (speed). . . . . . . . . . . . . . . . . . 296
10.4. DETECTION AND COUNTING DEVICES . . . . . . . . . . . . . . . 298
10.4.1. Survey meters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 298
10.4.2. Dose calibrator. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 299
10.4.3. Well counter. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 299
10.4.4. Intra-operative probes. . . . . . . . . . . . . . . . . . . . . . . . . . . 300
10.4.5. Organ uptake probe. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 302
10.5. QUALITY CONTROL OF DETECTION AND
COUNTING DEVICES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 305
10.5.1. Reference sources. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 305
10.5.2. Survey meter. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 306
10.5.3. Dose calibrator. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 307
10.5.4. Well counter. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 310
10.5.5. Intra-operative probe. . . . . . . . . . . . . . . . . . . . . . . . . . . . 310
10.5.6. Organ uptake probe. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311
CHAPTER 11. NUCLEAR MEDICINE IMAGING DEVICES. . . . . . . . . . 312

11.1. INTRODUCTION. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 312
11.2. GAMMA CAMERA SYSTEMS. . . . . . . . . . . . . . . . . . . . . . . . . 312
11.2.2. The Anger camera. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 314
11.2.3. SPECT. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 341
11.3. PET SYSTEMS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353
11.3.1. Principle of annihilation coincidence detection . . . . . . . 353
11.3.2. Design considerations for PET systems . . . . . . . . . . . . . 356
11.3.3. Detector systems. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 362
11.3.4. Data acquisition. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 369
11.3.5. Data corrections. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 380
11.4. SPECT/CT AND PET/CT SYSTEMS. . . . . . . . . . . . . . . . . . . . . 392
11.4.1. CT uses in emission tomography . . . . . . . . . . . . . . . . . . 392
11.4.2. SPECT/CT. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 393
11.4.3. PET/CT. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 394
CHAPTER 12. COMPUTERS IN NUCLEAR MEDICINE. . . . . . . . . . . . . . 398
12.1. PHENOMENAL INCREASE IN COMPUTING
CAPABILITIES. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 398
12.1.1. Moores law. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 398
12.1.2. Hardware versus peopleware . . . . . . . . . . . . . . . . . . . . 398
12.1.3. Future trends. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 399
12.2. STORING IMAGES ON A COMPUTER. . . . . . . . . . . . . . . . . . 400
12.2.1. Number systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 400
12.2.2. Data representation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 401
12.2.3. Images and volumes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 403
12.3. IMAGE PROCESSING. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 405
12.3.1. Spatial frequencies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 406
12.3.2. Sampling requirements. . . . . . . . . . . . . . . . . . . . . . . . . . 412
12.3.3. Convolution. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 412
12.3.4. Filtering. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 414
12.3.5. Band-pass filters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 416
12.3.6. Deconvolution. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 421
12.3.7. Image restoration filters. . . . . . . . . . . . . . . . . . . . . . . . . . 422
12.3.8. Other processing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 424
12.4. DATA ACQUISITION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 425
12.4.1. Acquisition matrix size and spatial resolution. . . . . . . . 426
12.4.2. Static and dynamic planar acquisition. . . . . . . . . . . . . . . 426
12.4.3. SPECT. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 427

12.4.4. PET acquisition. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 428
12.4.5. Gated acquisition. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 430
12.4.6. List-mode. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 431
12.5. FILE FORMAT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 431
12.5.1. File format design. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 432
12.5.2. Common image file formats. . . . . . . . . . . . . . . . . . . . . . 435
12.5.3. Movie formats. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 437
12.5.4. Nuclear medicine data requirements. . . . . . . . . . . . . . . . 437
12.5.5. Common nuclear medicine data storage formats . . . . . . 442
12.6. INFORMATION SYSTEM. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 443
12.6.1. Database . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 443
12.6.2. Hospital information system. . . . . . . . . . . . . . . . . . . . . . 445
12.6.3. Radiology information system . . . . . . . . . . . . . . . . . . . . 445
12.6.4. Picture archiving and communication system. . . . . . . . . 446
12.6.5. Scheduling. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 447
12.6.6. Broker . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 447
12.6.7. Security. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 447
CHAPTER 13. IMAGE RECONSTRUCTION. . . . . . . . . . . . . . . . . . . . . . . 449
13.1. INTRODUCTION. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 449
13.2. ANALYTICAL RECONSTRUCTION. . . . . . . . . . . . . . . . . . . . 450
13.2.1. Two dimensional tomography. . . . . . . . . . . . . . . . . . . . . 451
13.2.2. Frequencydistance relation. . . . . . . . . . . . . . . . . . . . . . 456
13.2.3. Fully 3D tomography. . . . . . . . . . . . . . . . . . . . . . . . . . . 457
13.2.4. Time of flight PET. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 466
13.3. ITERATIVE RECONSTRUCTION. . . . . . . . . . . . . . . . . . . . . . . 468
13.3.1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 468
13.3.2. Optimization algorithms. . . . . . . . . . . . . . . . . . . . . . . . . 473
13.3.3. Maximum-likelihood expectation-maximization . . . . . . 479
13.3.4. Acceleration. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 485
13.3.5. Regularization. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 488
13.3.6. Corrections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 495
13.4. NOISE ESTIMATION. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 507
13.4.1. Noise propagation in filtered back projection. . . . . . . . . 507
13.4.2. Noise propagation in maximum-likelihood
expectation-maximization. . . . . . . . . . . . . . . . . . . . . . . . 508
CHAPTER 14. NUCLEAR MEDICINE IMAGE DISPLAY. . . . . . . . . . . . . 512

14.1. INTRODUCTION. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 512
14.2. DIGITAL IMAGE DISPLAY AND VISUAL PERCEPTION. . . 513
14.2.1. Display resolution. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 514
14.2.2. Contrast resolution. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 515
14.3. DISPLAY DEVICE HARDWARE . . . . . . . . . . . . . . . . . . . . . . . 516
14.3.1. Display controller . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 516
14.3.2. Cathode ray tube. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 517
14.3.3. Liquid crystal display panel. . . . . . . . . . . . . . . . . . . . . . 519
14.3.4. Hard copy devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 521
14.4. GREY SCALE DISPLAY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 521
14.4.1. Grey scale standard display function. . . . . . . . . . . . . . . . 522
14.5. COLOUR DISPLAY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 525
14.5.1. Colour and colour gamut. . . . . . . . . . . . . . . . . . . . . . . . . 528
14.6. IMAGE DISPLAY MANIPULATION . . . . . . . . . . . . . . . . . . . . 530
14.6.1. Histograms. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 530
14.6.2. Windowing and thresholding. . . . . . . . . . . . . . . . . . . . . . 530
14.6.3. Histogram equalization. . . . . . . . . . . . . . . . . . . . . . . . . . 532
14.7. VISUALIZATION OF VOLUME DATA . . . . . . . . . . . . . . . . . . 533
14.7.1. Slice mode. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 533
14.7.2. Volume mode. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 534
14.7.3. Polar plots of myocardial perfusion imaging . . . . . . . . . 538
14.8. DUAL MODALITY DISPLAY. . . . . . . . . . . . . . . . . . . . . . . . . . 540
14.9. DISPLAY MONITOR QUALITY ASSURANCE. . . . . . . . . . . . 541
14.9.1. Acceptance testing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 542
14.9.2. Routine quality control . . . . . . . . . . . . . . . . . . . . . . . . . . 542
CHAPTER 15. DEVICES FOR EVALUATING IMAGING SYSTEMS. . . . 547
15.1. DEVELOPING A QUALITY MANAGEMENT SYSTEM
APPROACH TO INSTRUMENT QUALITY ASSURANCE. . . 547
15.1.1. Methods for routine quality assurance procedures. . . . . 547
15.2. HARDWARE (PHYSICAL) PHANTOMS. . . . . . . . . . . . . . . . . 550
15.2.1. Gamma camera phantoms. . . . . . . . . . . . . . . . . . . . . . . . 550
15.2.2. SPECT phantoms. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 558
15.2.3. PET phantoms. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 568
15.3. COMPUTATIONAL MODELS. . . . . . . . . . . . . . . . . . . . . . . . . . 575
15.3.1. Emission tomography simulation toolkits. . . . . . . . . . . . 577
15.4. ACCEPTANCE TESTING. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 578

15.4.1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 578
15.4.2. Procurement and pre-purchase evaluations. . . . . . . . . . . 580
15.4.3. Acceptance testing as a baseline for regular quality
assurance. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 583
15.4.4. What to do if the instrument fails acceptance testing. . . 584
15.4.5. Meeting the manufacturers specifications. . . . . . . . . . . 584
CHAPTER 16. FUNCTIONAL MEASUREMENTS IN NUCLEAR
MEDICINE. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 587
16.1. INTRODUCTION. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 587
16.2. NON-IMAGING MEASUREMENTS. . . . . . . . . . . . . . . . . . . . . 588
16.2.1. Renal function measurements. . . . . . . . . . . . . . . . . . . . . 588
16.2.2. 14C breath tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 591
16.3. IMAGING MEASUREMENTS. . . . . . . . . . . . . . . . . . . . . . . . . . 591
16.3.1. Thyroid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 592
16.3.2. Renal function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 594
16.3.3. Lung function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 596
16.3.4. Gastric function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 596
16.3.5. Cardiac function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 599
CHAPTER 17. QUANTITATIVE NUCLEAR MEDICINE. . . . . . . . . . . . . . 608
17.1. PLANAR WHOLE BODY BIODISTRIBUTION
MEASUREMENTS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 608
17.2. QUANTITATION IN EMISSION TOMOGRAPHY. . . . . . . . . 609
17.2.1. Region of interest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 609
17.2.2. Use of standard . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 610
17.2.3. Partial volume effect and the recovery coefficient. . . . . 610
17.2.4. Quantitative assessment . . . . . . . . . . . . . . . . . . . . . . . . . 612
17.2.5. Estimation of activity . . . . . . . . . . . . . . . . . . . . . . . . . . . 616
17.2.6. Evaluation of image quality. . . . . . . . . . . . . . . . . . . . . . 618
CHAPTER 18. INTERNAL DOSIMETRY. . . . . . . . . . . . . . . . . . . . . . . . . . . 621
18.1. THE MEDICAL INTERNAL RADIATION DOSE
FORMALISM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 621
18.1.1. Basic concepts. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 621
18.1.2. The time-integrated activity in the source region. . . . . . 626
18.1.3. Absorbed dose rate per unit activity (S value). . . . . . . . 628

18.1.4. Strengths and limitations inherent in the formalism. . . . 631
18.2. INTERNAL DOSIMETRY IN CLINICAL PRACTICE. . . . . . . 635
18.2.1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 635
18.2.2. Dosimetry on an organ level. . . . . . . . . . . . . . . . . . . . . . 636
18.2.3. Dosimetry on a voxel level. . . . . . . . . . . . . . . . . . . . . . . 637
CHAPTER 19. RADIONUCLIDE THERAPY. . . . . . . . . . . . . . . . . . . . . . . . 641
19.1. INTRODUCTION. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 641
19.2. THYROID THERAPIES. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 642
19.2.1. Benign thyroid disease . . . . . . . . . . . . . . . . . . . . . . . . . . 642
19.2.2. Thyroid cancer. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 643
19.3. PALLIATION OF BONE PAIN. . . . . . . . . . . . . . . . . . . . . . . . . . 645
19.3.1. Treatment specific issues. . . . . . . . . . . . . . . . . . . . . . . . . 646
19.4. HEPATIC CANCER. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 646
19.5. NEUROENDOCRINE TUMOURS. . . . . . . . . . . . . . . . . . . . . . . 647
19.6. NON-HODGKINS LYMPHOMA. . . . . . . . . . . . . . . . . . . . . . . 649
19.7. PAEDIATRIC MALIGNANCIES. . . . . . . . . . . . . . . . . . . . . . . . 650
19.7.1. Thyroid cancer. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 651
19.7.2. Neuroblastoma. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 651
19.8. ROLE OF THE PHYSICIST. . . . . . . . . . . . . . . . . . . . . . . . . . . . 652
19.9. EMERGING TECHNOLOGY. . . . . . . . . . . . . . . . . . . . . . . . . . . 654
19.10. CONCLUSIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 656
CHAPTER 20. MANAGEMENT OF THERAPY PATIENTS. . . . . . . . . . . . 658
20.1. INTRODUCTION. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 658
20.2. OCCUPATIONAL EXPOSURE . . . . . . . . . . . . . . . . . . . . . . . . . 658
20.2.1. Protective equipment and tools . . . . . . . . . . . . . . . . . . . 658
20.2.2. Individual monitoring . . . . . . . . . . . . . . . . . . . . . . . . . . . 659
20.3. RELEASE OF THE PATIENT. . . . . . . . . . . . . . . . . . . . . . . . . . . 659
20.3.1. The decision to release the patient. . . . . . . . . . . . . . . . . 660
20.3.2. Specific instructions for releasing the radioactive
patient . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 662
20.4. PUBLIC EXPOSURE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 665
20.4.1. Visitors to patients. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 665
20.4.2. Radioactive waste. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 665
20.5. RADIONUCLIDE THERAPY TREATMENT ROOMS

AND WARDS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 666
20.5.1. Shielding for control of external dose. . . . . . . . . . . . . . . 666
20.5.2. Designing for control of contamination . . . . . . . . . . . . . 668
20.6. OPERATING PROCEDURES. . . . . . . . . . . . . . . . . . . . . . . . . . . 668
20.6.1. Transport of therapy doses . . . . . . . . . . . . . . . . . . . . . . . 669
20.6.2. Administration of therapeutic radiopharmaceuticals. . . 669
20.6.3. Error prevention. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 670
20.6.4. Exposure rates and postings . . . . . . . . . . . . . . . . . . . . . . 670
20.6.5. Patient care in the treating facility. . . . . . . . . . . . . . . . . . 672
20.6.6. Contamination control procedures . . . . . . . . . . . . . . . . . 673
20.7. CHANGES IN MEDICAL STATUS. . . . . . . . . . . . . . . . . . . . . . 674
20.7.1. Emergency medical procedures. . . . . . . . . . . . . . . . . . . . 675
20.7.2. The radioactive patient in the operating theatre . . . . . . . 675
20.7.3. Radioactive patients on dialysis . . . . . . . . . . . . . . . . . . . 676
20.7.4. Re-admission of patients to the treating institution. . . . . 676
20.7.5. Transfer to another health care facility. . . . . . . . . . . . . . 677
20.8. DEATH OF THE PATIENT. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 677
20.8.1. Death of the patient following radionuclide therapy. . . . 678
20.8.2. Organ donation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 679
20.8.3. Precautions during autopsy. . . . . . . . . . . . . . . . . . . . . . . 679
20.8.4. Preparation for burial and visitation . . . . . . . . . . . . . . . . 680
20.8.5. Cremation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 681
APPENDIX I: ARTEFACTS AND TROUBLESHOOTING. . . . . . . . . . . . 684
APPENDIX II: RADIONUCLIDES OF INTEREST IN DIAGNOSTIC
AND THERAPEUTIC NUCLEAR MEDICINE . . . . . . . . . 719
ABBREVIATIONS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 723
SYMBOLS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 729
CONTRIBUTORS to drafting and review. . . . . . . . . . . . . . . . . . . . 735
CHAPTER 1
BASIC PHYSICS FOR NUCLEAR MEDICINE
E.B. PODGORSAK
Department of Medical Physics,
McGill University,
Montreal, Canada
A.L. KESNER
Division of Human Health,
International Atomic Energy Agency,
Vienna
P.S. SONI
Medical Cyclotron Facility,
Board of Radiation and Isotope Technology,
Bhabha Atomic Research Centre,
Mumbai, India
1.1. INTRODUCTION
The technologies used in nuclear medicine for diagnostic imaging have
evolved over the last century, starting with Rntgens discovery of X rays
and Becquerels discovery of natural radioactivity. Each decade has brought
innovation in the form of new equipment, techniques, radiopharmaceuticals,
advances in radionuclide production and, ultimately, better patient care. All
such technologies have been developed and can only be practised safely with
a clear understanding of the behaviour and principles of radiation sources and
radiation detection. These central concepts of basic radiation physics and nuclear
physics are described in this chapter and should provide the requisite knowledge
for a more in depth understanding of the modern nuclear medicine technology
discussed in subsequent chapters.
1.1.1. Fundamental physical constants
The chapter begins with a short list of physical constants of importance to
general physics as well as to nuclear and radiation physics. The data listed below
were taken from the CODATA set of values issued in 2006 and are available
1
CHAPTER 1
from a web site supported by the National Institute of Science and Technology in
Washington, DC, United States of America: http://physics.nist.gov/cuu/Constants
Avogadros number: NA=6.022 1023 mol1 or 6.022 1023 atoms/mol.
Speed of light in vacuum: c=2.998 108 m/s 3 108 m/s.
Electron charge: e=1.602 1019 C.
Electron and positron rest mass: me=0.511 MeV/c2.
Proton rest mass: mp=938.3 MeV/c2.
Neutron rest mass: mn=939.6 MeV/c2.
Atomic mass unit: u=931.5 MeV/c2.
Plancks constant: h=6.626 1034 J s.
Electric constant (permittivity of vacuum): 0=8.854 1012 C V1 m1.
Magnetic constant (permeability of vacuum): 0=4 107 V s A1 m1.
Newtonian gravitation constant: G=6.672 1011 m3 kg1 s2.
Proton mass/electron mass: mp/me=1836.0.
Specific charge of electron: e/me=1.758 1011 C/kg.
1.1.2. Physical quantities and units
A physical quantity is defined as a quantity that can be used in mathematical
equations of science and technology. It is characterized by its numerical value
(magnitude) and associated unit. The following rules apply to physical quantities
and their units in general:
Symbols for physical quantities are set in italics (sloping type), while
symbols for units are set in roman (upright) type (e.g. m=21 kg;
E=15 MeV; K=220 Gy).
Superscripts and subscripts used with physical quantities are set in italics if
they represent variables, quantities or running numbers; they are in roman
type if they are descriptive (e.g. Nx, m but max, Eab, tr).
Symbols for vector quantities are set in bold italics.
The currently used metric system of units is known as the International
System of Units (SI). The system is founded on base units for seven basic
physical quantities. All other quantities and units are derived from the seven base
quantities and units. The seven base SI quantities and their units are:
(a)
(b)
(c)
(d)
2
Length l: metre (m).

Mass m: kilogram (kg).
Time t: second (s).
Electric current I: ampere (A).
(e) Temperature T: kelvin (K).

(f) Amount of substance: mole (mol).
(g) Luminous intensity: candela (cd).
Examples of basic and derived physical quantities and their units are given
in Table 1.1.
TABLE1.1. BASIC QUANTITIES AND SEVERAL DERIVED PHYSICAL
QUANTITIES AND THEIR UNITS IN THE INTERNATIONAL SYSTEM OF
UNITS AND IN RADIATION PHYSICS
Physical
quantity
Units commonly
used in radiation
physics
Symbol SI unit
Conversion
Length
nm, , fm
1 m=109 nm=1010 =1015 fm
Mass
kg
MeV/c2
1 MeV/c2=1.78 1030 kg
Time
ms, s, ns, ps
1 s=103 ms=106 s=109 ns=1012 ps
Current
mA, A, nA, pA
1 A=103 mA=106 A=109 nA
Temperature
Mass density
kg/m3
Current density
A/m2
Velocity
m/s
Acceleration
m/s2
Frequency
Hz
Electric charge
Force
Pressure
Pa
Momentum
Ns
Energy
Power
T (in K)=T (in C) + 273.16

g/cm3
1 kg/m3=103 g/cm3
1 Hz=1 s1
e
1 e=1.602 1019 C
1 N=1 kg m s2
760 torr=101.3 kPa 1 Pa=1 N/m2 =7.5 103 torr

1 N s=1 kg m s1
eV, keV, MeV
1 eV=1.602 1019 J =103 keV

1 W=1 J/s=1 V A
CHAPTER 1
1.1.3. Classification of radiation

Radiation, the transport of energy by electromagnetic waves or atomic
particles, can be classified into two main categories depending on its ability to
ionize matter. The ionization potential of atoms, i.e. the minimum energy required
to ionize an atom, ranges from a few electronvolts for alkali elements to 24.6 eV
for helium which is in the group of noble gases. Ionization potentials for all other
atoms are between the two extremes.
Non-ionizing radiation cannot ionize matter because its energy per quantum
is below the ionization potential of atoms. Near ultraviolet radiation, visible
light, infrared photons, microwaves and radio waves are examples of
non-ionizing radiation.
Ionizing radiation can ionize matter either directly or indirectly because its
quantum energy exceeds the ionization potential of atoms. X rays, rays,
energetic neutrons, electrons, protons and heavier particles are examples of
ionizing radiation.
1.1.4. Classification of ionizing radiation
Ionizing radiation is radiation that carries enough energy per quantum to
remove an electron from an atom or a molecule, thus introducing a reactive and
potentially damaging ion into the environment of the irradiated medium. Ionizing
radiation can be categorized into two types: (i) directly ionizing radiation and
(ii) indirectly ionizing radiation. Both directly and indirectly ionizing radiation
can traverse human tissue, thereby enabling the use of ionizing radiation in
medicine for both imaging and therapeutic procedures.
Directly ionizing radiation consists of charged particles, such as electrons,
protons, particles and heavy ions. It deposits energy in the medium
through direct Coulomb interactions between the charged particle and
orbital electrons of atoms in the absorber.
Indirectly ionizing radiation consists of uncharged (neutral) particles
which deposit energy in the absorber through a two-step process. In the
first step, the neutral particle releases or produces a charged particle in
the absorber which, in the second step, deposits at least part of its kinetic
energy in the absorber through Coulomb interactions with orbital electrons
of the absorber in the manner discussed above for directly ionizing charged
particles.
1.1.5. Classification of indirectly ionizing photon radiation

Indirectly ionizing photon radiation consists of three main categories:
(i) ultraviolet, (ii) X ray and (iii) ray. Ultraviolet photons are of limited use in
medicine. Radiation used in imaging and/or treatment of disease consists mostly
of photons of higher energy, such as X rays and rays. The commonly accepted
difference between the two is based on the radiations origin. The term ray
is reserved for photon radiation that is emitted by the nucleus or from other
particle decays. The term X ray, on the other hand, refers to radiation emitted
by electrons, either orbital electrons or accelerated electrons (e.g. bremsstrahlung
type radiation).
With regard to their origin, the photons of the indirectly ionizing radiation
type fall into four categories: characteristic (fluorescence) X rays, bremsstrahlung
X rays, photons resulting from nuclear transitions and annihilation quanta.
1.1.6. Characteristic X rays
Orbital electrons have a natural tendency to configure themselves in such
a manner that they inhabit a minimal energy state for the atom. When a vacancy
is opened within an inner shell, as a result of an ionization or excitation process,
an outer shell electron will make a transition to fill the vacancy, usually within
a nanosecond for solid materials. The energy liberated in this transition may be
released in the form of a characteristic (fluorescence) photon of energy equal
to the difference between the binding energies of the initial and final vacancies.
Since different elements have different binding energies for their electronic
shells, the energy of the photon released in this process will be characteristic
of the particular atom. Rather than being emitted as a characteristic photon, the
transition energy may also be transferred to an orbital electron that is then emitted
with kinetic energy that is equal to the transition energy less the electron binding
energy. The emitted orbital electron is called an Auger electron.
1.1.7. Bremsstrahlung
The word bremsstrahlung can be translated from its original German
term as braking radiation, and is a name aptly assigned to the phenomenon.
When light charged particles (electrons and positrons) are slowed down or
negatively accelerated (decelerated) by interactions with other charged particles
in matter (e.g. by atomic nuclei), the kinetic energy that they lose is converted
to electromagnetic radiation, referred to as bremsstrahlung radiation. The energy
spectrum of bremsstrahlung is non-discrete (i.e. continuous) and ranges between
zero and the kinetic energy of the initial charged particle. Bremsstrahlung plays
5
CHAPTER 1
a central role in modern imaging and therapeutic equipment, since it can be

used to produce X rays on demand from an electrical energy source. The power
emitted in the form of bremsstrahlung photons is proportional to the square of the
particles charge and the square of the particles acceleration.
1.1.8. Gamma rays
When a nuclear reaction or spontaneous nuclear decay occurs, the process
may leave the product (daughter) nucleus in an excited state. The nucleus can
then make a transition to a more stable state by emitting a ray photon and the
process is referred to as decay. The energy of the photon emitted in decay is
characteristic of the nuclear energy transition, but the recoil of the emitting atom
produces a spectrum centred on the characteristic energy. Gamma rays typically
have energies above 100 keV and wavelengths less than 0.1 .
1.1.9. Annihilation quanta
When a parent nucleus undergoes plus decay or a high energy photon
interacts with the electric field of either the nucleus or the orbital electron, an
energetic positron may be produced. In moving through an absorber medium,
the positron loses most of its kinetic energy as a result of Coulomb interactions
with absorber atoms. These interactions result in collision loss when the
interaction is with an orbital electron of the absorber atom and in radiation loss
(bremsstrahlung) when the interaction is with the nucleus of the absorber atom.
Generally, after the positron loses all of its kinetic energy through collision and
radiation losses, it will undergo a final collision with an available orbital electron
(due to the Coulomb attractive force between the positively charged positron and
a local negatively charged electron) in a process called positron annihilation.
During annihilation, the positron and electron disappear and are replaced by
two oppositely directed annihilation quanta, each with an energy of 0.511 MeV.
This process satisfies a number of conservation laws: conservation of electric
charge, conservation of linear momentum, conservation of angular momentum
and conservation of total energy.
A percentage of positron annihilations occur before the positron expends all
of its kinetic energy and the process is then referred to as in-flight annihilation.
The two quanta emitted in in-flight annihilation are not of identical energies and
do not necessarily move in absolute opposite directions.
1.1.10. Radiation quantities and units

Accurate measurement of radiation is very important in all medical uses
of radiation, be it for diagnosis or treatment of disease. In diagnostic imaging
procedures, image quality must be optimized, so as to obtain the best possible
image with the lowest possible radiation dose to the patient to minimize the risk
of morbidity. In radiotherapy, the prescribed dose must be delivered accurately
and precisely to maximize the tumour control probability (TCP) and to minimize
the normal tissue complication probability (NTCP). In both instances, the risk
of morbidity includes acute radiation effects (radiation injury) as well as late
radiation-induced effects, such as induction of cancer and genetic damage.
Several quantities and units were introduced for the purpose of quantifying
radiation and the most important of these are listed in Table 1.2. Also listed are
the definitions for the various quantities and the relationships between the old
units and the SI units for these quantities. The definitions of radiation related
physical quantities are as follows:
Exposure X is related to the ability of photons to ionize air. Its unit, roentgen
(R), is defined as a charge of 2.58 104 coulombs produced per kilogram
of air.
Kerma K (acronym for kinetic energy released in matter) is defined for
indirectly ionizing radiation (photons and neutrons) as energy transferred to
charged particles per unit mass of the absorber.
Dose (also referred to as absorbed dose) is defined as energy absorbed per
unit mass of medium. Its SI unit, gray (Gy), is defined as 1 joule of energy
absorbed per kilogram of medium.
Equivalent dose HT is defined as the dose multiplied by a radiation
weighting factor wR. When different types of radiation are present, HT is
defined as the sum of all of the individual weighted contributions. The SI
unit of equivalent dose is the sievert (Sv).
Effective dose E of radiation is defined as the equivalent dose HT multiplied
by a tissue weighting factor wT. The SI unit of effective dose is also the
sievert (Sv).
Activity A of a radioactive substance is defined as the number of nuclear
decays per time. Its SI unit, becquerel (Bq), corresponds to one decay per
second.
CHAPTER 1
TABLE
1.2. RADIATION QUANTITIES, UNITS AND CONVERSION
BETWEEN OLD AND SI UNITS
Quantity
Definition
SI unit
10 4 C
kg air
Old unit
Conversion
10 4 C
kg air
Exposure X X =
Q
mair
2.58
Kerma K
K=
E tr
m
1 Gy = 1
J
kg
Dose D
D=
E ab
m
1 Gy = 1
J
kg
Equivalent
dose HT
HT=DwR
1 Sv
1 rem
1 Sv=100 rem
Effective
dose E
E = H Tw T
1 Sv
1 rem
1 Sv=100 rem
Activity A
A=lN
1 Bq = 1 s 1
1 Ci = 3.7 10 10 s 1
1 Bq =
1R=
1 esu
cm 3 airSTP
1 R = 2.58
1 rad = 100
erg
g
1 Gy=100 rad
1 Ci
3.7 10 10
1.2. BASIC DEFINITIONS FOR ATOMIC STRUCTURE

The constituent particles forming an atom are protons, neutrons and
electrons. Protons and neutrons are known as nucleons and form the nucleus of
the atom. Protons have a positive charge, neutrons are neutral and electrons have
a negative charge mirroring that of a proton. In comparison to electrons, protons
and neutrons have a relatively large mass exceeding the electron mass by a factor
of almost 2000 (note: mp/me=1836).
The following general definitions apply to atomic structure:
Atomic number Z is the number of protons and number of electrons in an
atom.
Atomic mass number A is the number of nucleons in an atom, i.e. the
number of protons Z plus the number of neutrons N in an atom: A=Z + N.
Atomic mass ma is the mass of a specific isotope expressed in atomic mass
units u, where 1 u is equal to one twelfth of the mass of the 12C atom (unbound,
at rest and in the ground state) or 931.5 MeV/c2. The atomic mass is smaller
than the sum of the individual masses of the constituent particles because of
the intrinsic energy associated with binding the particles (nucleons) within
the nucleus. On the other hand, the atomic mass is larger than the nuclear
mass M because the atomic mass includes the mass contribution of Z orbital
8
electrons while the nuclear mass M does not. The binding energy of orbital
electrons to the nucleus is ignored in the definition of the atomic mass.
While for 12C the atomic mass is exactly 12 u, for all other atoms ma does
not exactly match the atomic mass number A. However, for all atomic entities,
A (an integer) and ma are very similar to one another and often the same symbol
(A) is used for the designation of both. The mass in grams equal to the average
atomic mass of a chemical element is referred to as the mole (mol) of the
element and contains exactly 6.022 1023 atoms. This number is referred to as
the Avogadro constant NA of entities per mole. The atomic mass number of all
elements is, thus, defined such that A grams of every element contain exactly
NA atoms. For example, the atomic mass of natural cobalt is 58.9332 u. Thus,
one mole of natural cobalt has a mass of 58.9332 g and by definition contains
6.022 1023 entities (cobalt atoms) per mole of cobalt.
The number of atoms Na per mass of an element is given as:
Na NA
(1.1)
=
m
A
The number of electrons per volume of an element is:

Z
Na
N
N
= Z a = Z A (1.2)
V
m
A
The number of electrons per mass of an element is:

Z
Na
N
= Z A (1.3)
m
A
It should be noted that Z/A 0.5 for all elements with one notable exception
of hydrogen for which Z/A=1. Actually, Z/A slowly decreases from 0.5 for low
Z elements to 0.4 for high Z elements. For example, Z/A for 4He is 0.5, for 60Co is
0.45 and for 235U is 0.39.
If it is assumed that the mass of a molecule is equal to the sum of the masses
of the atoms that make up the molecule, then, for any molecular compound, there
are NA molecules per mole of the compound where the mole in grams is defined
as the sum of the atomic mass numbers of the atoms making up the molecule. For
example, 1 mole of water (H2O) is 18 g of water and 1 mole of carbon dioxide
(CO2) is 44 g of carbon dioxide. Thus, 18 g of water or 44 g of carbon dioxide
contain exactly NA molecules (or 3 NA atoms, since each molecule of water and
carbon dioxide contains three atoms).
9
CHAPTER 1
1.2.1. Rutherford model of the atom

At the beginning of the 20th century, the structure of the atom was not well
known. Scientific pioneers such as Dalton, Mendeleev and Thomson, among
others, were developing a common theory through their endeavours. Often noted
as a significant contribution to the modern understanding of the atom, is the
work performed by Rutherford and his colleagues Geiger and Marsden in 1909.
Through observation of the behaviour of positively charged particles traversing
a thin gold foil, Rutherford concluded that the positive charge and most of the
mass of the atom are concentrated in the atomic nucleus (diameter of a few
femtometres) and negative electrons are spread over the periphery of the atom
(diameter of a few ngstrms). This work was significant because it introduced a
new specialty of physics (nuclear physics) and demonstrated that the atom is not
simply a single particle, but instead is made up of smaller subatomic particles,
organized in an atom with well defined characteristics.
1.2.2. Bohr model of the hydrogen atom
Bohr expanded the Rutherford atomic model in 1913 using a set of four
postulates that combine classical, non-relativistic mechanics with the concept
of angular momentum quantization. The Bohr model of the atom can be said to
resemble a planetary model in that the protons and neutrons occupy a dense
central region called the nucleus and the electrons orbit the nucleus as planets
orbit the sun. The Bohr model introduces the concept that the angular momenta
of orbital electrons revolving around the nucleus in allowed orbits, radii of the
allowed electronic orbits (shells), velocities of orbital electrons in allowed orbits
and binding energies of orbital electrons in allowed orbits within the atom, are
restricted to certain discrete states. This means that angular momenta, radii,
velocities and binding energies of orbital electrons are quantized.
While scientific theory was later expanded to include the necessary
principles of quantum mechanics in our understanding of the atom, the Bohr
model is elegant and provides a simplistic, yet practical, view of the atom that is
still used for teaching atomic principles, and successfully deals with one-electron
entities, such as the hydrogen atom, the singly ionized helium atom and the
doubly ionized lithium atom.
1.3. BASIC DEFINITIONS FOR NUCLEAR STRUCTURE
According to the RutherfordBohr atomic model, most of the atomic mass
is concentrated in the atomic nucleus consisting of Z protons and (A Z) neutrons,
10
where Z is the atomic number and A the atomic mass number of a given nucleus.
In nuclear physics, the convention is to designate a nucleus X as ZA X , where A is
its atomic mass number and Z its atomic number; for example, the 60Co nucleus
226
is identified as 60
Ra nucleus as 226
27 Co and the
88 Ra. The atomic number Z is
often omitted in references to an atom because the atom is already identified by
its 13 letter symbol. In ion physics, the convention is to designate ions with +
or superscripts. For example, 42 He + stands for a singly ionized helium atom
and 42 He 2+ stands for a doubly ionized helium atom, also known as the particle.
With regard to relative values of atomic number Z and atomic mass number A of
nuclei, the following conventions apply:
An element may be composed of atoms that all have the same number of
protons, i.e. have the same atomic number Z, but have a different number of
neutrons (have different atomic mass numbers A). Such atoms of identical
Z but differing A are called isotopes of a given element.
The term isotope is often misused to designate nuclear species. For
example, 60Co, 137Cs and 226Ra are not isotopes, since they do not belong
to the same element. Rather than isotopes, they should be referred to as
nuclides. On the other hand, it is correct to state that deuterium (with a
nucleus called deuteron) and tritium (with a nucleus called triton) are heavy
isotopes of hydrogen or that 59Co and 60Co are isotopes of cobalt. Thus,
the term radionuclide should be used to designate radioactive species;
however, the term radioisotope is often used for this purpose.
A nuclide is an atomic species characterized by its nuclear composition (A, Z
and the arrangement of nucleons within the nucleus). The term nuclide
refers to all atomic forms of all elements. The term isotope is narrower
and only refers to various atomic forms of a given chemical element.
In addition to being classified into isotopic groups (common atomic
number Z), nuclides are also classified into groups with a common atomic mass
number A (isobars) and a common number of neutrons (isotones). For example,
60
67
Co and 60Ni are isobars with 60 nucleons each (A=60), and 67
31 Ga , 32 Ge and
3
4
67
33 As are isobars with atomic mass number 67, while H (tritium) and He are
isotones with two neutrons each (A Z=2), and 126 C, 137 N and 148 O are isotones
with six neutrons each.
A tool for remembering these definitions is as follows: isotopes have the
same number of protons Z; isotones have the same number of neutrons, A Z;
isobars have the same mass number A.
If a nucleus exists in an excited state for some time, it is said to be in
an isomeric (metastable) state. Isomers are, thus, nuclear species that have a
11
CHAPTER 1
common atomic number Z and a common atomic mass number A. For example,
99m
Tc is an isomeric state of 99Tc and 60mCo is an isomeric state of 60Co.
1.3.1. Nuclear radius
The radius R of a nucleus with atomic mass number A is estimated from the
following expression:
R = R0 3 A
(1.4)
where R0 is the nuclear radius constant equal to 1.25 fm. Since the range of A in
nature is from 1 to about 250, nuclear radius ranges from about 1 fm for a proton
to about 8 fm for heavy nuclei.
1.3.2. Nuclear binding energy
The sum of the masses of the individual components of a nucleus that
contains Z protons and (A Z) neutrons is larger than the actual mass of the
nucleus. This difference in mass is called the mass defect (deficit) m and its
energy equivalent mc2 is called the total binding energy EB of the nucleus. The
total binding energy EB of a nucleus can, thus, be defined as the energy liberated
when Z protons and (A Z) neutrons are brought together to form the nucleus.
The binding energy per nucleon (EB/A) in a nucleus (i.e. the total binding
energy of a nucleus divided by the number of nucleons in the given nucleus)
varies with the number of nucleons A and is of the order of ~8 MeV/nucleon.
A plot of the binding energy per nucleon EB/A in megaelectronvolts per
nucleon against the atomic mass number in the range from 1 to 250 is given in
Fig.1.1 and shows a rapid rise in EB/A at small atomic mass numbers, a broad
maximum of about 8.7 MeV/nucleon around A 60 and a gradual decrease in
EB/A at large A. The larger the binding energy per nucleon (EB/A) of an atom,
the larger is the stability of the atom. Thus, the most stable nuclei in nature are
the ones with A 60 (iron, cobalt, nickel). Nuclei of light elements (small A) are
generally less stable than nuclei with A 60, and the heaviest nuclei (large A) are
also less stable than nuclei with A 60.
12
(small A) are generally less stable than nuclei with A 60, and the heaviest nuclei (large A)
are also less stable than nuclei with A 60.
FIG.1.1. Binding energy per nucleon in megaelectronvolts per nucleon against atomic mass
FIG. 1.1.
Binding energy per nucleon in megaelectronvolts per nucleon against atomic mass
number A. Data are from the National Institute of Science and Technology (NIST).
number A. Data are from the National Institute of Science and Technology (NIST).
1.3.3. Nuclear fusion and fission
1.3.3. Nuclear fusion and fission
The peculiar
EB/A versus
A (Fig.
curve 1.1)
(Fig.1.1)
suggests
two for
The peculiar
shape shape
of theofEBthe
/A versus
A curve
suggests
two methods
methods
for
converting
mass
into
energy:
(i)
fusion
of
nuclei
at
low
A
converting mass into energy: (i) fusion of nuclei at low A and (ii) fission of nuclei at and
large A:
(ii) fission of nuclei at large A:
Fusion of two nuclei of very small mass, e.g. 21 H + 31 H 24 He + n , will
create a more massive nucleus and release a certain amount of energy.
Experiments using controlled nuclear fusion for production of energy have
so far not been successful in generating a net energy gain, i.e. the amount
of energy consumed is still larger than the amount created. However, fusion
remains an active field of research and it is reasonable to expect that in the
future controlled fusion will play an important role in the production of
electrical power.
Fission attained by bombardment of certain elements of large mass (such as
235
U) by thermal neutrons in a nuclear reactor will create two lower mass
and more stable nuclei, and transform some mass into kinetic energy of the
two product nuclei. Hahn, Strassman, Meitner and Frisch described fission
in 1939, and, in 1942, Fermi and colleagues at the University of Chicago
carried out the first controlled chain reaction based on nuclear fission.
13
CHAPTER 1
Since then, fission reactors have become an important means of production

of electrical power.
1.3.4. Two-particle collisions and nuclear reactions
A common problem in nuclear physics and radiation dosimetry is the
collision of two particles in which a projectile with mass m1, velocity 1 and
kinetic energy (EK)1 strikes a stationary target with mass m2 and velocity 2=0.
The probability or cross-section for a particular collision as well as the collision
outcome depends on the physical properties of the projectile (mass, charge,
velocity, kinetic energy) and the stationary target (mass, charge).
As shown schematically in Fig.1.2, the collision between the projectile
and the target in the most general case results in an intermediate compound that
subsequently decays into two reaction products: one of mass m3 ejected with
velocity 3 at an angle to the incident projectile direction, and the other of
mass m4 ejected with velocity 4 at an angle to the incident projectile direction.
Two-particle collisions are classified into three categories: (a) elastic
scattering, (b) inelastic collisions and (c) nuclear reactions:
(a) Elastic scattering is a special case of a two-particle collision in which the
products after the collision are identical to the products before collision,
i.e. m3=m1 and m4=m2, and the total kinetic energy and momentum
before the collision are equal to the total kinetic energy and momentum,
respectively, after the collision.
(b) In inelastic scattering of a projectile m1 on the target m2, similarly to elastic
scattering, the reaction products after collision are identical to the initial
products, i.e. m3=m1 and m4=m2; however, the incident projectile transfers
a portion of its kinetic energy to the target in the form of not only kinetic
energy but also intrinsic excitation energy E*.
(c) During a nuclear reaction, a collision between a projectile m1 and a target
m2 takes place and will result in the formation of two reaction products
m3 and m4, with the products having new atomic numbers. This process
is shown schematically in Fig.1.2. In any nuclear reaction, a number
of physical quantities must be conserved, most notably charge, linear
momentum and massenergy. In addition, the sum of atomic numbers Z
and the sum of atomic mass numbers A before and after the collision must
also be conserved.
14
The Q value of a nuclear reaction is defined as the difference between the

total rest energy before the reaction (m1c 2 + m 2c 2 ) and the total rest energy
after the reaction (m 3c 2 + m 4c 2 ) or:
Q = (m1c 2 + m 2c 2 ) (m 3c 2 + m 4c 2 ) (1.5)
Each Qtwo-particle
collision possesses a characteristic Q value(1.5)
that can
= (m c + m c ) - (m c + m c )
be either positive, zero or negative. For Q > 0, the collision is termed
Each two-particle collision possesses a characteristic Q value that can be either
exothermic
(also called exoergic) and results in a release of energy;
positive, zero or negative. For Q > 0, the collision is termed exothermic (also called
for Q=0,
the
is termed
elastic
Q <is termed
0, theelastic
collision is
exoergic)
andcollision
results in a release
of energy;
for Q = and
0, the for
collision
for Q < 0, the collision is termed endothermic (also called endoergic), and to
termedand
endothermic
(also
called
endoergic),
and
to
take
place,
it requires
take place, it requires an energy transfer from the projectile to the target. An
an energy
transfer
from
to while
the target.
An exothermic
reaction
exothermic
reaction
can the
occurprojectile
spontaneously,
an endothermic
reaction cannot
take place unless the projectile has kinetic energy exceeding the threshold energy
can occur
spontaneously,
while
an
endothermic
reaction
cannot
take
place
(EK)thr given as:
unless the projectile has kinetic energy exceeding the threshold energy
(EK)thr given as:
1
(m 3c 2 + m 4c 2 ) 2 - (m1c 2 + m 2c 2 ) 2
m1
(1.6)
(E ) 2 =
-Q1 +
m 2
c 2 c 2 + m c 2 ) 2
(m 3Kc thr
+ m 4c 2 ) 2 2m(2m
1 + m1 (1.6)
1
2
(E K ) thr =
m 2
2m 2c 2
2
2
2
and
restofenergies
ofm1the
projectile
m1,
wherewhere
m1c2m,1cm2, 2mc22c,2,m
4c theare
m33cc2 and
m4cm
are
rest the
energies
the projectile
, target
m2
products mproducts
target and
m2 reaction
and reaction
m3 and m4, respectively.
3 and m4, respectively.
FIG. 1.2. Schematic representation of a two-particle collision of a projectile (incident
particle)
of mass mrepresentation
a stationary target
with mass
velocity (incident
1 and velocity 1 striking
2
FIG.1.2.
Schematic
of a two-particle
collision
of ma2 and
projectile
= 0. An intermediate compound entity is formed temporarily that subsequently decays into
and
velocity
striking
a
stationary
target
with
mass
m
and
velocity
particle)twoofreaction
mass m
1
1 m4.
2
products
of mass m3 and
An RADIOACTIVITY
intermediate compound entity is formed temporarily that subsequently decays into
2=0. 1.4.
two reaction products of mass m3 and m4.
15
CHAPTER 1
1.4. RADIOACTIVITY
Radioactivity, also known as radioactive decay, nuclear decay, nuclear
disintegration and nuclear transformation, is a spontaneous process by which
an unstable parent nucleus emits a particle or electromagnetic radiation and
transforms into a more stable daughter nucleus that may or may not be stable.
The unstable daughter nucleus will decay further in a decay series until a stable
nuclear configuration is reached. Radioactive decay is usually accompanied by
emission of energetic particles or ray photons or both.
All radioactive decay processes are governed by the same general
formalism that is based on the definition of the activity A(t) and on a characteristic
parameter for each radioactive decay process, the radioactive decay constant l
with dimensions of reciprocal time, usually in s1. The main characteristics of
radioactive decay are as follows:
The radioactive decay constant multiplied by a time interval that is much
smaller than 1/ represents the probability that any particular atom of a
radioactive substance containing a large number N(t) of identical radioactive
atoms will decay (disintegrate) in that time interval. An assumption is made
that is independent of the physical environment of a given atom.
The activity A(t) of a radioactive substance containing a large number
N(t) of identical radioactive atoms represents the total number of decays
(disintegrations) per unit time and is defined as a product between N(t) and
, i.e.:
A (t ) = N (t ) (1.7)
The SI unit of activity is the becquerel (Bq) given as 1 Bq=1 s1. The
becquerel and hertz both correspond to s1, but hertz refers to the frequency of
periodic motion, while becquerel refers to activity.
The old unit of activity, the curie (Ci), was initially defined as the activity
of 1 g of 226Ra; 1 Ci @ 3.7 1010 s1.
Subsequently, the activity of 1 g of 226Ra was determined to be
3.665 1010 s1; however, the definition of the activity unit curie (Ci) was
kept as 1 Ci=3.7 1010 s1. Since the unit of activity the becquerel is 1 s1,
the SI unit becquerel (Bq) and the old unit curie (Ci) are related as follows:
1 Ci=3.7 1010 Bq and, consequently, 1 Bq=(3.7 1010)1 Ci=2.703 1011 Ci.
16
Specific activity a is defined as activity A per unit mass m, i.e.:

a=
A N N A
=
=

m
m
A
(1.8)
where NA is Avogadros number.

Specific activity a of a radioactive atom depends on the decay constant
and on the atomic mass number A of the radioactive atom. The units of specific
activity are Bq/kg (SI unit) and Ci/g (old unit).
1.4.1. Decay of radioactive parent into a stable or unstable daughter
The simplest form of radioactive decay involves a radioactive parent
nucleus P decaying with decay constant P into a stable or unstable daughter
nucleus D:
P
P
D (1.9)
The rate of depletion of the number of radioactive parent nuclei NP(t) is

equal to activity AP(t) at time t defined as the product N(t) in Eq.(1.7). We, thus,
have the following expression:
d N P (t )
= A P (t ) = P N P (t ) (1.10)
dt
The fundamental differential equation in Eq.(1.10) for NP(t) can be

rewritten in general integral form:
N P (t )
N P (0)
d N P (t )
=
NP
dt (1.11)
where NP(0) is the initial condition represented by the number of radioactive

nuclei at time t=0.
17
CHAPTER 1
Assuming that P is constant, Eq.(1.11) can be solved to obtain:

ln
N P (t )
= P t (1.12)
N P (0)
or
N P (t ) = N P (0)e P t (1.13)
Based on the definition of activity given in Eq.(1.7), the activity of parent

nuclei P at time t can be expressed as follows:
A P (t ) = P N P (t ) = P N P (0)e P t = A P (0)e P t
(1.14)
where AP(0)=PNP(0) is the initial activity of the radioactive substance.

The decay law of Eq.(1.14) applies to all radioactive nuclides irrespective
of their mode of decay; however, the decay constant P is different for each
parent radioactive nuclide P and is the most important defining characteristic of
a radioactive nuclide.
Two special time periods called half-life (T1/2)P and mean or average life P
are used to characterize a given radioactive parent substance P. The half-life (T1/2)P
of a radioactive substance P is the time during which the number of radioactive
nuclei of the substance decays to half of the initial value NP(0) present at time
t=0. It can also be stated that in the time of one half-life the activity AP(t) of a
radioactive substance decreases to one half of its initial value AP(0)=PNP(0):
1
N P [t = (T1/2 ) P ] = N P (0) = N P (0)e P (T1/2 ) P (1.15)
2
and
A P [t = (T1/2 ) P ] =
1
A (0) = A P (0)e P (T1/2 ) P (1.16)
2 P
From Eqs(1.15) and (1.16), it is noted that e P (T1/2 ) P must equal 1/2,
resulting in the following relationship between the decay constant P and half-life
(T1/2)P:
P =
18
ln 2
0.693
=
(1.17)
(T1/2 ) P (T1/2 ) P
Mean (average) life P of a radioactive parent P is defined as the time

required for the number NP of radioactive atoms or its activity AP to fall to
1/e=0.368 (or 36.8%) of the initial number of nuclei NP(0) or of the initial
activity AP(0), respectively. Thus, the following expressions describe the mean
half-life:
1
N P (t = P ) = N P (0) = 0.368 N P (0) = N P (0)e P P (1.18)
e
and
1
A P (t = P ) = A P (0) = 0.368 A P (0) = A P (0)e P P (1.19)
e
From Eqs(1.18) and (1.19), it is noted that e l P P must be equal to

1/e=e1=0.368, resulting in PP=1 and P=1/P. We now get the following
relationship between mean life P and half-life (T1/2)P using Eq.(1.17) and
P=1/P:
P =
ln 2
1

=
(T1/2 ) P P
(1.20)
and
P =
(T1/2 ) P
= 1.44(T1/2 ) P (1.21)
ln 2
A typical example of a radioactive decay for initial condition

AP(t=0)=AP(0) is shown in Fig.1.3 with a plot of parent activity AP(t) against
time t given in Eq.(1.14).
1.4.2. Radioactive series decay
The radioactive decay of parent P into stable daughter D, discussed in
Section 1.4.1, is the simplest known radioactive decay process; however, the
decay of a radioactive parent P with decay constant P into a radioactive (unstable)
daughter D which in turn decays with decay constant D into a stable or unstable
P
D
D
G ), is much more common and
grand-daughter G, i.e. ( P
results in a radioactive decay series for which the last decay product is stable.
The parent P in the decay series follows a straightforward radioactive
decay described by Eq.(1.16) for the rate of change of the number of parent
19
CHAPTER 1
FIG. 1.3. Activity A (t) plotted against time t for a simple decay of a radioactive parent P into
FIG. 1.3. Activity APP(t) plotted against time t for a simple decay of a radioactive parent P into
a stable or unstable daughter D. The concepts of half-life (T1/2)P and mean life P are also
a stable or unstable daughter D. The concepts of half-life (T1/2)P and mean life P are also
illustrated. The
The area
area under
illustrated.
under the
the exponential
exponential decay
decaycurve
curvefrom
fromtt==0 0totot t==isisequal
equaltotothethe
(0)
where
A
(0)
is
the
initial
activity
of
the
parent
P.
The
slope
of
the
product
A
product APP(0)PP where APP(0) is the initial activity of the parent P. The slope of thetangent
tangenttoto
A
(0)
and
this
tangent
crosses
the
abscissa
axis
at
t=
the
decay
curve
at
t=0
is
equal
to
P
P
at Pt. =
the decay curve at t = 0 is equal to PAP(0) and this tangent crosses the abscissa axis
P.
nuclei dNP(t)/dt. The rate of change of the number of daughter nuclei dND(t)/dt,
however, is more complicated and consists of two components, one being the
1.4.2. Radioactive series decay
supply of new daughter nuclei D through the decay of P given as PNP(t) and
the other
being the
lossPofdecay
daughter
nucleidaughter
D fromD,the
decay of
D to G 1.4.1,
givenisasthe
Radioactive
parent
into stable
discussed
in Section
simplest
known
radioactive
process;
however, the
of a radioactive parent P with
DND(t),
resulting
in thedecay
following
expression
fordecay
dND(t)/dt:
decay constant P into a radioactive (unstable) daughter D which in turn decays with decay
l
l
constant D into a stable or unstable granddaughter G, i.e. ( P
D G ), is much
dN D (t )
P t
more common
and
inDa(tradioactive
decay series
forD (which
the last decay product is
) DN
) = P N P (0)e
t ) (1.22)
= PN
DN
P (tresults
dt
stable.
The parent P in the decay series follows a straightforward radioactive decay described
by
Eq. the
(1.16)
for the
rate of change
of the
number
of parent
dNP(t)/dt.
rate of
assuming
thatnuclei
(i) the
initialThe
number
With
initial
conditions
for time
t=0
change
of the
number
of daughter
nuclei dN
however,
is are
morenocomplicated
and
consists
D(t)/dt,
(0),
and
(ii)
there
daughter
D
nuclei
of parent
nuclei
P is
NP(t=0)=N
P
of two components, one being the supply of new daughter nuclei D through the decay of P
present, i.e. ND(t=0)=0, the solution of the differential equation in Eq.(1.22)
given as PNP(t) and the other being the loss of daughter nuclei D from the decay of D to G
readsasasfollows:
given
DND(t), resulting in the following expression for dND(t)/dt:
P
d N (t )
D
t
= l P N P (tP) - l D eND(Ptt)
= leP
- l D N D (t )
(1.22)
NPD(0)e
N D (t ) =dN
(0)
(1.23)
t P

-l P t
With the initial conditions for time t = 0 assuming that (i) the initial number of parent nuclei P
is NP(t = 0) = NP(0), and (ii) there are no daughter D nuclei present, i.e. ND(t = 0) = 0, the
solution of the differential equation in Eq. (1.22) reads as follows:
20
Recognizing that the activity of the daughter AD(t) is DND(t), the daughter
activity AD(t) is written as:
D P P t
D P t
e D t = A P (0)
e D t
e
e
D P
D P

D
1 P t
e D t = A P (t )
=A P (0)
e
1 e ( D P )t
P
D P
1
D
A D (t ) = N P (0)
(1.24)
where
AD(t) is the activity at time t of the daughter nuclei equal to DND(t);
AP(0) is the initial activity of the parent nuclei present at time t=0;
and AP(t) is the activity at time t of the parent nuclei equal to PNP(t).
While for initial conditions AP(t=0)=AP(0) and AD(t=0)=0, the parent P
activity AP(t) follows the exponential decay law of Eq.(1.14) shown in Fig.1.3,
the daughter D activity AD(t) starts at 0, then initially rises with time t, reaches
a maximum at a characteristic time t=(tmax)D, and then diminishes to reach 0 at
t=. The characteristic time (tmax)D is given as follows:
P
D
(1.25)
(t max )D =
P D
ln
1.4.3. Equilibrium in parentdaughter activities

In many parent P daughter D grand-daughter G relationships, after a
certain time t the parent and daughter activities reach a constant ratio independent
of a further increase in time t. This condition is referred to as radioactive
equilibrium and can be analysed by examining the behaviour of the activity ratio
AD(t)/AP(t) obtained from Eq.(1.24) as:
A D (t )
D
1
1 e ( D P )t =
1 e ( D P )t (1.26)
=
P
A P (t ) D P
1
21
CHAPTER 1
Three possibilities merit special consideration:

(a) The half-life of the daughter exceeds that of the parent: (T1/2)D > (T1/2)P
resulting in D < P. The activity ratio AD(t)/AP(t) of Eq.(1.26) is written as:
A D (t )
D ( P D ) t
e
=
1 (1.27)
A P (t ) P D
The ratio AD(t)/AP(t) increases exponentially with time t, indicating that no

equilibrium between the parent activity AP(t) and daughter activity AD(t)
will be reached.
(b) The half-life of the daughter is shorter than that of the parent: (T1/2)D < (T1/2)P
or D > P.
The activity ratio AD(t)/AP(t) at large t becomes a constant equal to
D/(D P) and is then independent of t and larger than unity, implying
transient equilibrium, i.e.:
A D (t )
D
=
= const > 1
A P (t ) D P
(1.28)
(c) The half-life of the daughter is much shorter than that of the parent:
(T1/2)D (T1/2)P or D P.
For relatively large time t tmax, the activity ratio AD(t)/AP(t) of Eq.(1.28)
simplifies to:
A D (t )
1 (1.29)
A P (t )
The activity of the daughter AD(t) very closely approximates that of its parent
AP(t), i.e. AD(t) AP(t), and they decay together at the rate of the parent. This
special case of transient equilibrium in which the daughter and parent activities
are essentially identical is called secular equilibrium.
1.4.4. Production of radionuclides (nuclear activation)
In 1896, Henri Becquerel discovered natural radioactivity, and in 1934
Frdric Joliot and Irne Curie-Joliot discovered artificial radioactivity. Most
natural radionuclides are produced through one of four radioactive decay chains,
each chain fed by a long lived and heavy parent radionuclide. The vast majority
of currently known radionuclides, however, are human-made and artificially
produced through a process of nuclear activation which uses bombardment of a
22
stable nuclide with a suitable energetic particle or high energy photons to induce
a nuclear transformation. Various particles or electromagnetic radiation generated
by a variety of machines are used for this purpose, most notably neutrons from
nuclear reactors for neutron activation, protons from cyclotrons or synchrotrons
for proton activation, and X rays from high energy linear accelerators for nuclear
photoactivation.
Neutron activation is important in production of radionuclides used
for external beam radiotherapy, brachytherapy, therapeutic nuclear medicine
and nuclear medicine imaging also referred to as molecular imaging; proton
activation is important in production of positron emitters used in positron
emission tomography (PET) imaging; and nuclear photoactivation is important
from a radiation protection point of view when components of high energy
radiotherapy machines become activated during patient treatment and pose a
potential radiation risk to staff using the equipment.
A more in depth discussion of radionuclide production can be found in
Chapter 4.
1.4.5. Modes of radioactive decay
Nucleons are bound together to form the nucleus by the strong nuclear
force that, in comparison to the protonproton Coulomb repulsive force, is at
least two orders of magnitude larger but of extremely short range (only a few
femtometres). To bind the nucleons into a stable nucleus, a delicate equilibrium
between the number of protons and the number of neutrons must exist. For light
(low A) nuclear species, a stable nucleus is formed by an equal number of protons
and neutrons (Z=N). Above the nucleon number A 40, more neutrons than
protons must constitute the nucleus to form a stable configuration in order to
overcome the Coulomb repulsion among the charged protons.
If the optimal equilibrium between protons and neutrons does not exist, the
nucleus is unstable (radioactive) and decays with a specific decay constant into
a more stable configuration that may also be unstable and decay further, forming
a decay chain that eventually ends with a stable nuclide.
Radioactive nuclides, either naturally occurring or artificially produced
by nuclear activation or nuclear reactions, are unstable and strive to reach
more stable nuclear configurations through various processes of spontaneous
radioactive decay that involve transformation to a more stable nuclide and
emission of energetic particles. General aspects of spontaneous radioactive decay
may be discussed using the formalism based on the definitions of activity A and
decay constant without regard for the actual microscopic processes that underlie
the radioactive disintegrations.
23
CHAPTER 1
A closer look at radioactive decay processes shows that they are divided
into six categories, consisting of three main categories of importance to
medical use of radionuclides and three categories of less importance. The main
categories are: (i) alpha () decay, (ii) beta () decay encompassing three related
decay processes (beta minus, beta plus and electron capture) and (iii) gamma
() decay encompassing two competing decay processes (pure decay and
internal conversion). The three less important radioactive decay categories are:
(i) spontaneous fission, (ii) proton emission decay and (iii) neutron emission
decay.
Nuclides with an excess number of neutrons are referred to as neutron-rich;
nuclides with an excess number of protons are referred to as proton-rich. The
following features are notable:
For a slight protonneutron imbalance in the nucleus, radionuclides decay
by decay characterized by transformation of a proton into a neutron in
+ decay, and transformation of a neutron into a proton in decay.
For a large protonneutron imbalance in the nucleus, the radionuclides
decay by emission of nucleons: particles in decay, protons in proton
emission decay and neutrons in neutron emission decay.
For very large atomic mass number nuclides (A > 230), spontaneous fission,
which competes with decay, is also possible.
Excited nuclei decay to their ground state through decay. Most of these
transformations occur immediately upon production of the excited state by either
or decay; however, a few exhibit delayed decays that are governed by their
own decay constants and are referred to as metastable states (e.g. 99mTc).
Nuclear transformations are usually accompanied by emission of energetic
particles (charged particles, neutral particles, photons, etc.). The particles released
in the various decay modes are as follows:
Alpha particles in decay;
Electrons in decay;
Positrons in + decay;
Neutrinos in + decay;
Antineutrinos in decay;
Gamma rays in decay;
Atomic orbital electrons in internal conversion;
Neutrons in spontaneous fission and in neutron emission decay;
Heavier nuclei in spontaneous fission;
Protons in proton emission decay.
24
In each nuclear transformation, a number of physical quantities must

be conserved. The most important of these quantities are: (i) total energy,
(ii) momentum, (iii) charge, (iv) atomic number and (v) atomic mass number
(number of nucleons).
The total energy of particles released by the transformation process is
equal to the net decrease in the rest energy of the neutral atom, from parent P to
daughter D. The disintegration (decay) energy, often referred to as the Q value
for the radioactive decay, is defined as follows:
Q = {M (P) [M (D) + m]}c 2 (1.30)
where M(P), M(D) and m are the nuclear rest masses (in unified atomic mass
units u) of the parent, daughter and emitted particles, respectively.
For radioactive decay to be energetically possible, the Q value must be
greater than zero. This means that spontaneous radioactive decay processes
release energy and are called exoergic or exothermic. For Q > 0, the energy
equivalent of the Q value is shared as kinetic energy between the particles emitted
in the decay process and the daughter product. Since the daughter generally has a
much larger mass than the other emitted particles, the kinetic energy acquired by
the daughter is usually negligibly small.
1.4.6. Alpha decay
In decay, a radioactive parent nucleus P decays into a more stable
daughter nucleus D by ejecting an energetic particle. Since the particle is a
4
He nucleus ( 42 He 2+ ), in decay the parents atomic number Z decreases by two
and its atomic mass number A decreases by four:
A
A4
4
2+
Z P Z 2 d + 2 he
A4
Z 2 d + a
(1.31)
Naturally occurring particles have kinetic energies between 4 and 9 MeV;

their range in air is between 1 and 10 cm, and their range in tissue is between
10 and 100 m.
25
CHAPTER 1
Typical examples of decay are the decay of 226Ra with a half-life of

1602 a into 222Rn which is also radioactive and decays with a half-life of 3.82 d
into 218Po:
226
222
88 Ra
86 Rn + a
T =1602 a
1/2
and
222
218
86 Rn
84 Po + a
T = 3.82 d
1/2
(1.32)
1.4.7. Beta minus decay

In beta minus () decay, a neutron-rich parent nucleus P transforms
a neutron into a proton and ejects an electron e and an electronic
antineutrino e . Thus, in decay, the atomic number of the daughter increases
by one, i.e. ZD=ZP + 1, the atomic mass number remains constant, i.e. AD=AP,
and the general relationship for decay is given as:
A
A
Z P Z +1 D + e + e (1.33)
A typical example of decay is the decay of 60Co with a half-life of

60
*
5.26 a into an excited state of 60Ni ( 60
28 Ni ). Excited states of Ni progress to the
60
12
ground state of Ni instantaneously (within 10 s) through emission of rays
in decay:
60
*
60
27 Co
28 Ni
T = 5.26 a
1/2
+ e + e (1.34)
1.4.8. Beta plus decay

In beta plus (+) decay, a proton-rich parent nucleus P transforms a proton
into a neutron and ejects a positron e+ and an electronic neutrino e . Thus, in +
decay, the atomic number of the daughter decreases by one, i.e. ZD=ZP 1, the
atomic mass number, just as in decay, remains constant, i.e. AD =AP, and the
general relationship for + decay is written as:
A
A
+
Z P Z 1 D + e
+ e (1.35)
Radionuclides undergoing + decay are often called positron emitters and

are used in medicine for functional imaging with the special imaging technique
26
PET. The most common tracer for PET studies is fluorodeoxyglucose (FDG)
labelled with 18F which serves as a good example of + decay:
18
18
+
9 F
8O+e
T =110 min
1/2
+ e (1.36)
1.4.9. Electron capture

Electron capture radioactive decay may occur when an atomic electron
ventures inside the nuclear volume, is captured by a proton, triggers a proton to
neutron transformation, and an ejection of an electronic neutrino e. In electron
capture, as in + decay, the atomic number of the daughter decreases by one,
i.e. ZD=ZP 1, and its atomic mass number, just as in and + decay, remains
constant, i.e. AD = AP. The general relationship for electron capture decay is
written as:
A
Z P+e
Z 1 D + e (1.37)
T =60 d
1/2
A simple example of electron capture decay is the decay of 125I with a

half-life of 60 d into an excited state of 125Te which then decays to the ground
state of 125Te through decay and internal conversion:
125
53 I + e
125
*
52 Te + e
T =60 d
1/2
(1.38)
1.4.10. Gamma decay and internal conversion

Alpha decay as well as the three decay modes (, + and electron capture)
may produce a daughter nucleus in an excited state without expending the full
amount of the decay energy available. The daughter nucleus will then reach its
ground state, either instantaneously or with some time delay (isomeric metastable
state), through one of the following two processes:
(a) By emitting the excitation energy in the form of one or more photons in a
decay process referred to as decay.
(b) By transferring the excitation energy to one of its associated atomic
orbital electrons (usually a K shell electron) in a process called internal
conversion. The vacancy left behind by the ejected orbital electron is filled
by a transition from a higher atomic shell, resulting in characteristic X rays
and/or Auger electrons.
27
CHAPTER 1
In most radioactive or decays, the daughter nucleus de-excitation

occurs instantaneously (i.e. within 1012 s), so that the emitted rays are referred
to as if they were produced by the parent nucleus (e.g. 60Co rays). The rays
produced from isomeric transitions are attributed to the isomeric daughter
product (e.g. 99mTc rays).
The decay process and the internal conversion process may be represented,
respectively, as follows:
A *
ZX
A
ZX+g
A +
ZX
(1.39)
and
A *
ZX
+ e
A
ZX
(1.40)
where
A *
ZX
stands for an excited state of the nucleus
and ZA X + is the singly ionized state of atom

decay.
A
Z X;
A
ZX
following internal conversion
*
An example of decay is the transition of an excited 60
28 Ni nucleus,
60
60
resulting from the decay of 27 Co , into stable 28 Ni through an emission of two
rays with energies of 1.17 and 1.33 MeV. An example of internal conversion
*
decay is the decay of excited 125
52 Te which results from an electron capture decay
of 125I into stable 125
52 Te through emission of 35 keV rays (7 %) and internal
conversion electrons (93%).
1.4.11. Characteristic (fluorescence) X rays and Augerelectrons

A large number of radionuclides used in nuclear medicine (e.g. 99mTc, 123I,
Tl, 64Cu) decay by electron capture and/or internal conversion. Both processes
leave the atom with a vacancy in an inner atomic shell, most commonly the
K shell. The vacancy in the inner shell is filled by an electron from a higher level
atomic shell and the binding energy difference between the two shells is either
emitted as a characteristic X ray (fluorescence photon) or transferred to a higher
shell orbital electron which is then emitted from the atom as an Auger electron
with a kinetic energy equal to the transferred energy minus the binding energy
of the emitted Auger electron. Emission of characteristic photons and Auger
electrons is discussed further in Section 1.6.4.
201
28
1.5. ELECTRON INTERACTIONS WITH MATTER

As an energetic charged particle, such as an electron or positron, traverses
an absorbing medium, it experiences a large number of Coulomb interactions
with the nuclei and orbital electrons of absorber atoms before its kinetic energy is
expended. In each interaction, the charged particles path may be altered (elastic
or inelastic scattering) and it may lose some of its kinetic energy that will be
transferred to the medium or to photons. Charged particle interactions with orbital
electrons of the absorber result in collision (ionization loss); interactions with
nuclei of the absorber result in radiation loss. Each of these possible interactions
between the charged particle and absorber atom is characterized by a specific
cross-section (probability) for the particular interaction. The energy loss of the
charged particle propagating through an absorber depends on the properties of the
particle, such as its mass, charge, velocity and energy, as well as on the properties
of the absorber, such as its density and atomic number.
Stopping power is a parameter used to describe the gradual loss of energy
of the charged particle as it penetrates into an absorbing medium. Two classes
of stopping power are known: collision stopping power scol results from charged
particle interaction with orbital electrons of the absorber and radiation stopping
power srad results from charged particle interaction with nuclei of the absorber.
The total stopping power stot is the sum of the collision stopping power and the
radiation stopping power.
1.5.1. Electronorbital interactions
Coulomb interactions between the incident electron or positron and orbital
electrons of an absorber result in ionizations and excitations of absorber atoms.
Ionization is described as ejection of an orbital electron from the absorber atom,
thereby producing an ion. Excitation, on the other hand, is defined as the transfer
of an orbital electron of the absorber atom from an allowed orbit to a higher
allowed orbit (shell), thereby producing an excited atom. Atomic excitations and
ionizations result in collision energy loss and are characterized by collision (also
known as ionization) stopping powers.
1.5.2. Electronnucleus interactions
Coulomb interactions between the incident electron or positron and nuclei
of the absorber atom result in particle scattering. The majority of these scattering
events are elastic and result in no energy loss. However, when the scattering is
inelastic, the incident charged particle loses part of its kinetic energy through
production of X ray photons referred to as bremsstrahlung radiation. This energy
29
CHAPTER 1
loss is characterized by radiation stopping powers and is governed by the Larmor

relationship which states that the rate of energy loss is proportional to the square
of the particles acceleration and the square of the particles charge.
1.6. PHOTON INTERACTIONS WITH MATTER
1.6.1. Exponential absorption of photon beam in absorber
The most important parameter used in characterization of X ray or ray
penetration into absorbing media is the linear attenuation coefficient . This
coefficient depends on the energy h of the photon and the atomic number Z of
the absorber, and may be described as the probability per unit path length that
a photon will have an interaction with the absorber. The attenuation coefficient
is determined experimentally by aiming a narrowly collimated monoenergetic
photon beam h onto a suitable radiation detector and placing an absorber
material of varying thickness x between the photon source and the detector. The
absorber decreases the detector signal intensity from I(x) which is measured with
no absorber in the beam (x=0) to I(x) measured with an absorber of thickness
x > 0 in the beam.
dI(x)/dx, the rate of change in beam intensity I(x) transmitted through an
absorber of thickness x, is equal to the product of the attenuation coefficient
and the beam intensity I(x) at thickness x (seeEq.(1.41)). Alternatively, it can
be said that an absorber of thickness dx reduces the beam intensity by dI and the
fractional reduction in intensity dI/I is equal to the product of the attenuation
coefficient and the absorber layer thickness dx (seeEq.(1.42)). The following
expressions are obtained, respectively:
dI (x)
= I (x) (1.41)
dx
and
dI
= dx (1.42)
I
where the negative sign is used to indicate a decrease in signal I(x) with an
increase in absorber thickness x.
It should be noted that Eqs(1.41) and (1.42) can be considered identical.
30
The form of Eq.(1.41) is identical to the form of Eq.(1.10) that deals with
simple radioactive decay; however, it must be noted that in radioactive decay the
product N(t) is defined as activity A(t), while in photon beam attenuation the
product I(x) does not have a special name and symbol.
Integration of Eq.(1.42) over absorber thickness x from 0 to x and over
intensity I(x) from the initial intensity I(0) (no absorber) to intensity I(x) at
absorber thickness x, gives:
I ( x)
I (0)
dI
=
I
dx (1.43)
0
resulting in:
I ( x) = I (0)e x (1.44)
where it is assumed that in a homogeneous absorber the attenuation coefficient

is uniform and independent of absorber thickness x.
1.6.2. Characteristic absorber thicknesses
Equation (1.44) represents the standard expression for the exponential
attenuation of a monoenergetic narrow photon beam. A typical exponential plot
of intensity I(x) against absorber thickness x of Eq.(1.44) is shown in Fig.1.4 for
a monoenergetic and narrow photon beam. The figure also defines three special
absorber thicknesses used for characterization of photon beams: half-value layer
(HVL), mean free path (MFP) and tenth-value layer (TVL):
HVL (or x1/2) is defined as the thickness of a homogeneous absorber that
attenuates the narrow beam intensity I(0) to one half (50%) of the original
intensity, i.e. I(x1/2)=0.5I(0). The relationship between the HVL x1/2 and
the attenuation coefficient is determined from the basic definition of the
HVL as follows:
I ( x 1/2 ) = 0.5I (0) = I (0)e x 1/2 (1.45)
resulting in:
1
= e x 1/2
2
or
x 1/2 = ln 2 = 0.693
31
CHAPTER 1
and
HVL = x 1/2 =
ln 2
(1.46)
MFP (or x ) or relaxation length is the thickness of a homogeneous

absorber that attenuates the beam intensity I(0) to 1/e=0.368 (36.8%) of
its original intensity, i.e. I( x )=0.368I(0). The photon MFP is the average
distance a photon of energy h travels through a given absorber before
undergoing an interaction. The relationship between the MFP x and the
attenuation coefficient is determined from the basic definition of the
MFP as follows:
1
I ( x ) = I (0) = 0.368 I (0) = I (0)e x (1.47)
e
resulting in:
1
= e x
e
or
x = 1
and
MFP = x =
1
(1.48)
TVL (or x1/10) is the thickness of a homogeneous absorber that attenuates

the beam intensity I(0) to one tenth (10%) of its original intensity, i.e.
I(x1/10)=0.1I(0). The relationship between the TVL x1/10 and the attenuation
coefficient is determined from the basic definition of the TVL as follows:
I ( x 1/10 ) = 0.1I (0) = I (0)e x 1/10 (1.49)
resulting in:
1
= e x 1/10
10
32
or
x 1/10 = ln 10 = 2.303
and
TVL = x 1/10 =
ln 10
(1.50)
From Eqs(1.46), (1.48) and (1.50), the linear attenuation coefficient may
be expressed in terms of x1/2, x and x1/10, respectively, as follows:
=
ln 2 1 ln 10
(1.51)
= =
x 1/2 x x 1/10
resulting in the following relationships among the characteristic thicknesses:

x 1/2 = (ln 2)x =
ln 2
x
= 0.301x 1/10 (1.52)
ln 10 1/10
e
e
e
FIG.1.4. Intensity I(x) against absorber thickness x for a monoenergetic photon beam.
Half-value layer x1/2, mean free path x and tenth-value layer x1/10 are also illustrated. The
area under the exponential attenuation curve from x=0 to x= is equal to the product I(0) x
where I(0) is the initial intensity of the monoenergetic photon beam. The slope of the tangent
to the attenuation curve at x=0 is equal to I(0) and this tangent crosses the abscissa (x) axis
at x = x .
33
CHAPTER 1
1.6.3. Attenuation coefficients

In addition to the linear attenuation coefficient , three other related
attenuation coefficients are in use for describing photon beam attenuation
characteristics in absorbers: mass attenuation coefficient m, atomic attenuation
coefficient a and electronic attenuation coefficient e . The attenuation
coefficients are related as follows:
= m = n a = Z n e (1.53)
where

is the mass density of the absorber;

is the number of atoms Na per volume V of the absorber,
n
i.e. n = N a / V , and N a / V = N a / m = N A / A with m the mass of the
absorber, NA Avogadros number of atoms per mole and A the atomic mass
of the absorber in grams per mole;
Z is the atomic number of the absorber;

and Z n is the number of electrons per unit volume of absorber,

i.e. Z n = ZN A / A.
In radiation dosimetry, two energy-related coefficients are in use: (i) the
energy transfer coefficient tr that accounts for the mean energy transferred E tr
from photons to charged particles (electrons and positrons) in a photonatom
interaction; and (ii) the energy absorption coefficient ab that accounts for the
mean energy absorbed E ab in the medium. The two coefficients are given as
follows:
tr =
E tr
h
(1.54)
E ab

h
(1.55)
and
ab =
34
The light charged particles (electrons and positrons) released or produced

in the absorbing medium through various photon interactions will either:
(a) Deposit their energy to the medium through Coulomb interactions with
orbital electrons of the absorbing medium (collision loss also referred to as
ionization loss); or
(b) Radiate their kinetic energy away in the form of photons through Coulomb
interactions with the nuclei of the absorbing medium (radiation loss).
Typical examples of the mass attenuation coefficient / are shown in
Fig.1.5 with plots of / against photon energy h (in solid dark curves) for
carbon and lead in the energy range from 0.001 to 1000 MeV. Carbon with Z=6
is an example of a low Z absorber; lead with Z=82 is an example of a high
Z absorber. Comparing the two absorbers, it can be noted that at intermediate
photon energies (around 1 MeV), carbon and lead have a similar / of about
0.1 cm2/g. On the other hand, at low photon energies, the / of lead significantly
exceeds the / of carbon, and at energies above 10 MeV, the / of carbon is
essentially flat while the / of lead increases with increasing energy.
1.6.4. Photon interactions on the microscopic scale
The general trends in / depicted in Fig.1.5 reflect the elaborate
dependence of / on the energy h of the photon and the atomic number Z of
the absorber. In penetrating an absorbing medium, photons may experience
various interactions with the atoms of the medium. On a microscopic scale, these
interactions involve either the nuclei of the absorbing medium or the orbital
electrons of the absorbing medium:
(a) Photon interactions with the nucleus of the absorber atoms may be direct
photonnucleus interactions (photonuclear reaction) or interactions
between the photon and the electrostatic field of the nucleus (nuclear pair
production);
(b) Photon interactions with orbital electrons of absorber atoms are
characterized as interactions between the photon and either a loosely bound
electron (Compton effect, triplet production) or a tightly bound electron
(photoelectric effect, Rayleigh scattering).
A loosely bound electron is an electron whose binding energy EB is much
smaller in comparison with the photon energy h, i.e. EB h. An interaction
between a photon and a loosely bound electron is considered to be an interaction
between a photon and a free (i.e. unbound) electron.
35
CHAPTER 1
Carbon
Lead
FIG.1.5. Mass attenuation coefficient / against photon energy h in the range from 1 keV
FIG. 1.5. Mass attenuation coefficient / against photon energy h in the range from 1 keV
to 1000 MeV for carbon (a) and lead (b). In addition to the total coefficients /, the individual
to 1000 MeV for carbon (a) and lead (b). In addition to the total coefficients /, the
coefficients
the photoelectric
Rayleigheffect,
scattering,
Compton
scattering
and pair
individualfor
coefficients
for the effect,
photoelectric
Rayleigh
scattering,
Compton
scattering
production
(including
triplet
production)
are
also
shown.
Data
are
from
the
National
and pair production (including triplet production) are also shown. Data Institute
are from the
ofNational
Science and
Technology
(NIST).
Institute
of Science
and Technology (NIST).
As far as the photon fate after the interaction with an atom is concerned, there are two
possible outcomes: (i) the photon disappears and is absorbed completely (photoelectric effect,
A tightly
bound triplet
electron
is an electron
whose
binding
energy
EB is
is
nuclear
pair production,
production,
photonuclear
reaction)
and (ii)
the photon
comparable
larger its
than
or slightly
smaller
than (Rayleigh
the photon
energy or
h.loses
For part
a of
scattered andto,changes
direction
but keeps
its energy
scattering)
its
energy
(Compton
effect).
photon interaction to occur with a tightly bound electron, the binding energy
The most
with smaller
atoms of
the the
absorber
EB of the electron
mustimportant
be of thephoton
order interactions
of, but slightly
than,
photonare: the
Compton effect, photoelectric effect, nuclear pair production, electronic pair production
energy,
i.e. EB h. An
interaction between a photon and a tightly bound electron
(triplet production)
and photonuclear reactions. In some of these interactions, energetic
iselectrons
considered
an
interaction
photon(photoelectric
and the atomeffect,
as a whole.
are released frombetween
absorberaatoms
Compton effect, triplet
production) and electronic vacancies are left in absorber atoms; in other interactions, a portion
of the incident photon energy is used to produce free electrons and positrons. All of these
36
As far as the photon fate after the interaction with an atom is concerned,
there are two possible outcomes: (i) the photon disappears and is absorbed
completely (photoelectric effect, nuclear pair production, triplet production,
photonuclear reaction) and (ii) the photon is scattered and changes its direction
but keeps its energy (Rayleigh scattering) or loses part of its energy (Compton
effect).
The most important photon interactions with atoms of the absorber are:
the Compton effect, photoelectric effect, nuclear pair production, electronic pair
production (triplet production) and photonuclear reactions. In some of these
interactions, energetic electrons are released from absorber atoms (photoelectric
effect, Compton effect, triplet production) and electronic vacancies are left in
absorber atoms; in other interactions, a portion of the incident photon energy
is used to produce free electrons and positrons. All of these light charged
particles move through the absorber and either deposit their kinetic energy in the
absorber (dose) or transform part of it back into radiation through production of
bremsstrahlung radiation.
The fate of electronic vacancies produced in photon interactions with
absorber atoms is the same as the fate of vacancies produced in electron capture
and internal conversion. As alluded to in Section 1.4.11, an electron from a higher
atomic shell of the absorber atom fills the electronic vacancy in a lower shell and
the transition energy is emitted either in the form of a characteristic X ray (also
called a fluorescence photon) or an Auger electron and this process continues
until the vacancy migrates to the outer shell of the absorber atom. A free electron
from the environment will eventually fill the outer shell vacancy and the absorber
ion will revert to a neutral atom in the ground state.
A vacancy produced in an inner shell of an absorber atom migrates to
the outer shell and the migration is accompanied by emission of a series of
characteristic photons and/or Auger electrons. The phenomenon of emission
of Auger electrons from an excited atom is called the Auger effect. Since each
Auger transition converts an initial single electron vacancy into two vacancies,
a cascade of low energy Auger electrons is emitted from the atom. These low
energy electrons have a very short range in tissue but may produce ionization
densities comparable to those produced in an particle track.
The branching between a characteristic photon and an Auger electron is
governed by the fluorescence yield which, as shown in Fig.1.6, for a given
electronic shell, gives the number of fluorescence photons emitted per vacancy
in the shell. The fluorescence yield can also be defined as the probability of
emission of a fluorescence photon for a given shell vacancy. Consequently, as
also shown in Fig.1.6, (1 ) gives the probability of emission of an Auger
electron for a given shell vacancy.
37
Auger electrons. The phenomenon of emission of Auger electrons from an excited atom is
called the Auger effect. Since each Auger transition converts an initial single electron vacancy
into two vacancies, a cascade of low energy
Auger 1electrons is emitted from the atom. These
CHAPTER
low energy electrons have a very short range in tissue but may produce ionization densities
comparable to those produced in an particle track.
FIG.1.6. Fluorescence yields , and M, against atomic number Z of the absorber. Also
FIG. 1.6. Fluorescence yields KK, LL and M
, against atomic number Z of the absorber. Also
shown are
are probabilities
probabilities for
shown
for the
the Auger
Auger effect,
effect, given
given as
as (1
(1 ).
). Data
Data are
arefrom
fromthe
theNational
National
Institute
of
Science
and
Technology
(NIST).
Institute of Science and Technology (NIST).
The branching between a characteristic photon and an Auger electron is governed by
the
fluorescence
yield which,
1.6.5. Photoelectric
effect as shown in Fig. 1.6, for a given electronic shell, gives the
number of fluorescence photons emitted per vacancy in the shell. The fluorescence yield
can also be defined as the probability of emission of a fluorescence photon for a given shell
In Consequently,
the photoelectric
called
the photoeffect),
the
photonof
vacancy.
as alsoeffect
shown(sometimes
in Fig. 1.6, (1
) gives
the probability of
emission
an
Auger
electron
for
a
given
shell
vacancy.
interacts with a tightly bound orbital electron of an absorber atom, the photon
disappears and the orbital electron is ejected from the atom as a so-called
photoelectron, with a kinetic energy EK given as:
1.6.5. Photoelectric effect
In the photoelectric effect (sometimes called the photoeffect), the photon

interacts
bound orbital electron of an absorber atom, the photon disappears and
(1.56)
E K = hwith
EaBtightly
where
h is the incident photon energy;
and EB is the binding energy of the ejected photoelectron.
A general diagram of the photoelectric effect is provided (seeFig.1.9(a)).
For the photoelectric effect to happen, the photon energy h must exceed
the binding energy EB of the orbital electron to be ejected and, moreover, the
closer h is to EB, the higher the probability of the photoelectric effect happening.
The photoelectric mass attenuation coefficient / is plotted in Fig.1.5 for carbon
and lead as one of the grey curves representing the components of the total /
attenuation coefficient. The sharp discontinuities in the energy h are called
38
absorption edges and occur when h becomes equal to the binding energy EB of a
given atomic shell. For example, the K absorption edge occurs at h=88 keV in
lead, since the K shell binding energy EB in lead is 88 keV. Absorption edges for
carbon occur at h < 1 keV and, thus, do not appear in Fig.1.5(a).
As far as the photoelectric attenuation coefficient dependence on photon
energy h and absorber atomic number Z is concerned, the photoelectric atomic
attenuation coefficient a goes approximately as Z5/(h)3, while the photoelectric
mass attenuation coefficient / goes approximately as Z4/(h)3.
As evident from Fig.1.5, the photoelectric attenuation coefficient /
is the major contributor to the total attenuation coefficient / at relatively low
photon energies where h is of the order of the K shell binding energy and less
than 0.1 MeV. At higher photon energies, first the Compton effect and then pair
production become the major contributors to the photon attenuation in the absorber.
1.6.6. Rayleigh (coherent) scattering
In Rayleigh scattering (also called coherent scattering), the photon
interacts with the full complement of tightly bound atomic orbital electrons of
an absorber atom. The event is considered elastic in the sense that the photon
loses essentially none of its energy h but is scattered through a relatively
small scattering angle . A general diagram of Rayleigh scattering is given
(seeFig.1.9(b)).
Since no energy transfer occurs from photons to charged particles, Rayleigh
scattering plays no role in the energy transfer attenuation coefficient and energy
absorption coefficient; however, it contributes to the total attenuation coefficient
/ through the elastic scattering process. The Rayleigh atomic attenuation
coefficient aR is proportional to Z2/(h)2 and the Rayleigh mass attenuation
coefficient R/ is proportional to Z/(h)2.
As a result of no energy transfer from photons to charged particles in the
absorber, Rayleigh scattering is of no importance in radiation dosimetry. As far
as photon attenuation is concerned, however, the relative importance of Rayleigh
scattering in comparison to other photon interactions in tissue and tissue
equivalent materials amounts to only a few per cent of the total / but it should
not be neglected.
1.6.7. Compton effect (incoherent scattering)
The Compton effect (also called incoherent scattering or Compton
scattering) is described as an interaction between a photon and a free as well
as stationary electron. Of course, the interacting electron is not free, rather it is
bound to a nucleus of an absorbing atom, but the photon energy h is much larger
39
CHAPTER 1
than the binding energy EB of the electron (EB h), so that the electron is said
to be loosely bound or essentially free and stationary.
In the Compton effect, the photon loses part of its energy to the recoil
(Compton) electron and is scattered as a photon h' through a scattering angle ,
as shown schematically in Fig.1.7. In the diagram, the interacting electron is at
the origin of the Cartesian coordinate system and the incident photon is oriented
in the positive direction along the abscissa (x) axis. The scattering angle is the
angle between the direction of the scattered photon h' and the positive abscissa
axis while the recoil angle is the angle between the direction of the recoil
electron and the positive abscissa axis. A general diagram of the Compton effect
is given (seeFig.1.9(c)).
FIG.1.7. Schematic diagram of the Compton effect in which an incident photon of energy
h=1 MeV interacts with a free and stationary electron. A photon with energy h=0.505 MeV
is produced and scattered with a scattering angle =60.
Considerations of conservation of energy and momentum result in the

following three equations for the Compton effect:
(a) Conservation of energy:
h + mec 2 = h '+ mec 2 + E K
40
or
h = h '+ E K
(1.57)
(b) Conservation of momentum on the abscissa (x) axis:

pv =
h '
cos +
c
me
1
cos (1.58)
c2
(c) Conservation of momentum on the ordinate (y) axis:

0=
h '
sin +
c
me
1
sin (1.59)
c2
where
mec2 is the rest energy of the electron (0.511 MeV);
EK is the kinetic energy of the recoil (Compton) electron;
is the velocity of the recoil (Compton) electron;
and c is the speed of light in a vacuum (3 108 m/s).
From equations describing conservation of energy (Eq.(1.57)) and
conservation of momentum (Eqs(1.58) and (1.59)), the basic Compton equation
(also referred to as the Compton wavelength-shift equation) can be derived and is
expressed as follows:
' = =
h
(1 cos ) = C (1 cos ) (1.60)
m ec
where
is the wavelength of the incident photon (c/);
is the wavelength of the scattered photon (c/ );
is the wavelength shift in Compton effect ( );
and c, defined as c = h /(mec) = 2 c /(mec 2 ) = 0.024 , is the so-called Compton
wavelength of the electron.
From the Compton equation (Eq.(1.60)), it is easy to show that the
scattered photon energy h and the recoil electron kinetic energy EK depend
41
CHAPTER 1
on the incident photon energy h as well as on the scattering angle and are,
respectively, given as:
h '(h , ) = h
1
(1.61)
1 + (1 cos )
and
E KC (h , ) = h h ' = h h
1
(1 cos )
(1.62)
= h
1 + (1 cos )
1 + (1 cos )
where is the incident photon energy h normalized to electron rest energy mec2,
i.e. = h /(mec 2 ).
Using Eq.(1.61), it is easy to show that energies of forward-scattered
photons (=0), side-scattered photons ( = /2) and backscattered photons
(=) are in general given as follows:
h ' =0 = h (1.63)
h ' = =
2
h

1+
(1.64)
and
h ' = =
h
(1.65)
1 + 2
For very large incident photon energies (h ), they are given as:
h ' =0 = h
h ' = = mec 2
2
(1.66)

(1.67)
and
h ' = =
42
m ec 2
(1.68)
2
From the conservation of momentum equations (Eqs(1.58) and (1.59)),

the following expression for the relationship between the scattering angle and
recoil electron angle can be derived:
cot = (1 + )tan
(1.69)
and
tan =
cot

1+
2
(1.70)
Since the range of is from 0 (forward-scattering) through /2 (sidescattering) to (backscattering), it is noted that the corresponding range of is
from =/2 at =0 through to =(1 + )1 for =/2 to =0 at =.
The Compton electronic attenuation coefficient e C steadily decreases
with increasing h from a theoretical value of 0.665 1024 cm2/electron (known
as the Thomson cross-section) at low photon energies to 0.21 1024 cm2/electron
at h =1 MeV, 0.51 1024 cm2/electron at h=10 MeV, and
0.008 1024 cm2/electron at h=100 MeV.
Since Compton interaction is a photon interaction with a free electron, the
Compton atomic attenuation coefficient a C depends linearly on the absorber
atomic number Z, while the electronic coefficient e C and the mass coefficient
C/ are essentially independent of Z. This independence of Z can be observed in
Fig.1.5, showing that C/ for carbon (Z=6) and lead (Z=82) at intermediate
photon energies (~1 MeV), where Compton effect predominates, are equal to
about 0.1 cm2/electron irrespective of Z.
Equation (1.62) gives the energy transferred from the incident photon
to the recoil electron in the Compton effect as a function of the scattering
angle . The maximum energy transfer to recoil electron occurs when the photon
is backscattered (=) and the Compton maximum energy transfer fraction
(fC)max is then given as:
( f C ) max =
(E KC ) max
2
=
(1.71)
h
1 + 2
The mean energy transfer in the Compton effect f C is determined by

normalizing (to incident photon energy h) the mean energy transferred to the
Compton electron E KC . This quantity is very important in radiation dosimetry
and is plotted against incident photon energy h in the Compton graph presented
in Fig.1.8. The figure shows that the fractional energy transfer to recoil electrons
43
h in the Compton graph presented in Fig. 1.8. The figure shows that the fractional energy
transfer to recoil electrons is quite low at low photon energies ( f C = 0.02 at h = 0.01 MeV)
and then slowly rises through f C = 0.44 atCHAPTER
h = 1 MeV
to reach f C = 0.80 at h = 100 MeV
1
and approaches one asymptotically at very high incident photon energies.
FIG.1.8. Maximum and mean fractions of incident photon energy h transferred to the recoil
FIG. 1.8. Maximum

meaneffect.
fractions
offrom
incident
photon
energy
h transferred
to the recoil
electron in theand
Compton
Data are
the National
Institute
of Science
and Technology
electron in
the Compton effect. Data are from the National Institute of Science and
(NIST).
Technology (NIST).
1.6.8. Pair production
is quite low at low photon energies ( f C = 0.02 at h=0.01 MeV) and then slowly
risesthe
through
f C =photon
0.44 atenergy
h=1 hMeV
to reach
at h=100
MeV
exceeds
2mfeCc2= =0.80
1.022
MeV, with
meand
c2 being the
When
incident
approaches
one asymptotically
at verythe
highproduction
incident photon
rest energy
of the electron
and positron,
of anenergies.
electronpositron pair in
conjunction with a complete absorption of the incident photon by the absorber atom becomes
1.6.8.
Pair production
energetically
possible.
For the effect to occur, three quantities must be conserved: energy,
charge and momentum. To conserve the linear momentum simultaneously with total energy
incident
photon
energy
h itexceeds
2meoccur
c2=1.022
with electric
and charge, theWhen
effectthe
cannot
occur
in free
space;
can only
in theMeV,
Coulomb
2
the rest (atomic
energy of
the electron
and positron,
field of a mcollision
nucleus
or orbital
electron)the
thatproduction
can take ofupana suitable
ec being partner
pair in conjunction
withTwo
a complete
absorption
of the incident
fraction ofelectronpositron
the momentum carried
by the photon.
types of
pair production
are known:
photon by the absorber atom becomes energetically possible. For the effect to
occur, three quantities must be conserved: energy, charge and momentum. To
conserve the linear momentum simultaneously with total energy and charge,
the effect cannot occur in free space; it can only occur in the Coulomb electric
field of a collision partner (atomic nucleus or orbital electron) that can take up
a suitable fraction of the momentum carried by the photon. Two types of pair
production are known:
If the collision partner is an atomic nucleus of the absorber, the pair
production event is called nuclear pair production and is characterized by
a photon energy threshold slightly larger than two electron rest masses
(2mec2=1.022 MeV).
44
Less
probable,
but is
nonetheless
possible,ofisthe
pair
production
in the
Coulomb

If the
collision
partner
an atomic nucleus
absorber,
the pair
production
event
is called
nuclear
pair production
is characterized
by aThe
photon
energy
threshold
field of
an orbital
electron and
of an
absorber atom.
event
is called
slightly
larger than
electron or
resttriplet
massesproduction
(2mec2 = 1.022
electronic
pair two
production
and MeV).
its threshold photon
2nonetheless possible, is pair production in the Coulomb field of an
Less
probable,
but
energy is 4mec =2.044 MeV.
orbital electron of an absorber atom. The event is called electronic pair production or
triplet production and its threshold photon energy is 4mec2 = 2.044 MeV.
FIG.1.9. Schematic diagrams of the most important modes of photon interaction with atoms of
FIG. 1.9. Schematic diagrams of the most important modes of photon interaction with atoms
an absorber: (a) photoelectric effect; (b) Rayleigh scattering; (c) Compton effect; (d) nuclear
of an
absorber: (a) photoelectric effect; (b) Rayleigh scattering; (c) Compton effect; (d)
pair production; and (e) electronic pair production (triplet production).
nuclear
pair production; and (e) electronic pair production (triplet production).
The two pair production attenuation coefficients, despite having different origins,
45 are
usually dealt with together as one parameter referred to as pair production. The component
that the nuclear pair production contributes usually exceeds 90%. Nuclear pair production and
CHAPTER 1
The two pair production attenuation coefficients, despite having different

origins, are usually dealt with together as one parameter referred to as pair
production. The component that the nuclear pair production contributes usually
exceeds 90%. Nuclear pair production and electronic pair/triplet production are
shown schematically in Figs 1.9(d) and (e), respectively.
The probability of pair production is zero for photon energy below the
threshold
value and
increasesare
rapidly
with
photon energy
the threshold.
electronic
pair/triplet
production
shown
schematically
in above
Figs 1.9(d)
and 1.9(e),
The pair production atomic attenuation coefficient a and the pair production
respectively.
Theattenuation
probability coefficient
of pair production
is zero
for photon energy
threshold value
mass
/ vary
approximately
as Z2 below
and Z,therespectively,
and where
increases
rapidly
with
photon
energy
above
the
threshold.
The
pair
production
atomic
Z is the atomic number of the absorber.
attenuation coefficient a k and the pair production mass attenuation coefficient / vary
Z, respectively,of
where
Z is the effects
atomic number of the absorber.
approximately
as Z2 andpredominance
1.6.9. Relative
individual
1.6.9. Relative predominance of individual effects
As is evident from the discussion above, photons have several options for
interaction
with from
absorber
atoms. Five
of the
most have
important
As is evident
the discussion
above,
photons
severalphoton
optionsinteractions
for interaction
withare
absorber
Five of theinmost
important
photon
schematically
Fig.1.9.
Nuclear
andinteractions
electronic are
pairshown
production
are
shownatoms.
schematically
in Fig.
1.9. Nuclear
andand
electronic
production
are usually
combined and treated under the
usually
combined
treatedpair
under
the header
pair production.
header pair
production.
The
probability for a photon to undergo any one of the various interaction
The probability for a photon to undergo any one of the various interaction phenomena
phenomena with an absorber depends on the energy h of the photon and the
with an absorber depends on the energy h of the photon and the atomic number Z of the
atomic number Z of the absorber. In general, the photoelectric effect predominates
absorber. In general, the photoelectric effect predominates at low photon energies, the
Compton
effect
at intermediate
energies
and pair
at low
photon
energies, the
Compton
effect
at intermediate
energies
and pair
production
at high photon
energies.
production at high photon energies.
Representation of the relative predominance of the three main processes of

FIG.FIG.1.10.
1.10. Representation
of the relative predominance of the three main processes of photon
photon
interaction
an absorber
the photoelectriceffect,
effect, Compton
Compton effect
pairpair
interaction with an with
absorber
atom: atom:
the photoelectric
effectand
and
production.
production.
46 Figure 1.10 shows the regions of relative predominance of the three most important
individual effects with h and Z as parameters. The two curves display the points on the (h,
Z) diagram for which C = at low photon energies and for which C = for high photon
energies and, thus, delineate regions of photoelectric effect predominance at low photon
energies, Compton effect predominance at intermediate photon energies and pair production
Figure 1.10 shows the regions of relative predominance of the three most
important individual effects with h and Z as parameters. The two curves display
the points on the (h, Z) diagram for which C= at low photon energies and for
which C= for high photon energies and, thus, delineate regions of photoelectric
effect predominance at low photon energies, Compton effect predominance at
intermediate photon energies and pair production predominance at high photon
energies. Figure 1.10 also indicates how the regions of predominance are affected
by the absorber atomic number. For example, a 100 keV photon will interact with
a lead absorber (Z=82) predominantly through the photoelectric effect and with
soft tissue (Zeff 7.5) predominantly through the Compton effect. A 10 MeV
photon, on the other hand, will interact with lead predominantly through pair
production and with tissue predominantly through the Compton effect.
1.6.10. Macroscopic attenuation coefficients
For a given photon energy h and absorber atomic number Z, the
macroscopic attenuation coefficient and energy transfer coefficient tr are
given as a sum of coefficients for individual photon interactions discussed above
(photoelectric, Rayleigh, Compton and pair production):
=
NA
( + a R + a C + a)
A a
(1.72)
and
tr =
NA
N
+ ( a C ) tr + a tr = A a f PE + a C f C + a f PP (1.73)
A a tr
A
where all parameters are defined in sections dealing with the individual
microscopic effects.
It should be noted that in Rayleigh scattering there is no energy transfer to
charged particles.
The energy absorption coefficient ab (often designated en in the literature)
is derived from tr of Eq.(1.73) as follows:
ab = en = tr (1 g ) (1.74)
where g is the mean radiation fraction accounting for the fraction of the
mean energy transferred from photons to charged particles and subsequently
lost by charged particles through radiation losses. These losses consist of two
47
CHAPTER 1
components: the predominant bremsstrahlung loss and the small, yet not always
negligible, in-flight annihilation loss.
1.6.11. Effects following photon interactions with absorber
and summary of photon interactions
In the photoelectric effect, Compton effect and triplet production, vacancies
are produced in atomic shells of absorber atoms through the ejection of orbital
electrons from atomic shells. For the diagnostic range and megavoltage range
of photons used for diagnosis and treatment of disease with radiation, the shell
vacancies occur mainly in inner atomic shells and are followed by characteristic
radiation or Auger electrons, the probability of the former given by fluorescence
yield (seeFig.1.6).
Pair production and triplet production are followed by the annihilation
of the positron with a free electron producing two annihilation quanta, most
commonly with an energy of 0.511 MeV each and emitted at 180 from each
other to satisfy conservation of energy, momentum and charge.
BIBLIOGRAPHY
ATTIX, F.H., Introduction to Radiological Physics and Radiation Dosimetry, Wiley, New York
(1986).
CHERRY, S.R., SORENSON, J.A., PHELPS, M.E., Physics in Nuclear Medicine, 3rd edn,
Saunders, Philadelphia, PA (2003).
EVANS, R.D., The Atomic Nucleus, Krieger Publishing, Malabar, FL (1955).
HENDEE, W., RITENOUR, E.R., Medical Imaging Physics, 4th edn, Wiley, New York (2002).
JOHNS, H.E., CUNNINGHAM, J.R., The Physics of Radiology, 3rd edn, Thomas, Springfield,
IL (1984).
KHAN, F., The Physics of Radiation Therapy, 4th edn, Lippincott, Williams and Wilkins,
Baltimore, MD (2009).
KRANE, K., Modern Physics, 3rd edn, Wiley, New York (2012).
PODGORSAK, E.B., Radiation Physics for Medical Physicists, 2nd edn, Springer, Heidelberg,
New York (2010).
ROHLF, J.W., Modern Physics from to Z0, Wiley, New York (1994).
48
CHAPTER 2
BASIC RADIOBIOLOGY
R.G. DALE
Department of Surgery and Cancer,
Faculty of Medicine,
Imperial College London,
London, United Kingdom
J. WONDERGEM*
Division of Human Health,
Vienna
2.1. INTRODUCTION
Radiobiology is the study (both qualitative and quantitative) of the actions of
ionizing radiations on living matter. Since radiation has the ability to cause changes
in cells which may later cause them to become malignant, or bring about other
detrimental functional changes in irradiated tissues and organs, consideration of
the associated radiobiology is important in all diagnostic applications of radiation.
Additionally, since radiation can lead directly to cell death, consideration of the
radiobiological aspects of cell killing is essential in all types of radiation therapy.
2.2. RADIATION EFFECTS AND TIMESCALES
At the microscopic level, incident rays or particles may interact with
orbital electrons within the cellular atoms and molecules to cause excitation or
ionization. Excitation involves raising a bound electron to a higher energy state,
but without the electron having sufficient energy to leave the host atom. With
ionization, the electron receives sufficient energy to be ejected from its orbit and
to leave the host atom. Ionizing radiations (of which there are several types) are,
thus, defined through their ability to induce this electron ejection process, and
Present address: Department of Radiology, Leiden University Medical Centre, Leiden,

Netherlands.
49
CHAPTER 2
the irradiation of cellular material with such radiation gives rise to the production
of a flux of energetic secondary particles (electrons). These secondary particles,
energetic and unbound, are capable of migrating away from the site of their
production and, through a series of interactions with other atoms and molecules,
give up their energy to the surrounding medium as they do so.
This energy absorption process gives rise to radicals and other chemical
species and it is the ensuing chemical interactions involving these which are the
true causatives of radiation damage. Although the chemical changes may appear to
operate over a short timescale (~105s), this period is nonetheless a factor of ~1018
longer than the time taken for the original particle to traverse the cell nucleus. Thus,
on the microscopic scale, there is a relatively long period during which chemical
damage is inflicted (Table2.1).
It is important to note that, irrespective of the nature of the primary
radiation (which may be composed of particles and/or electromagnetic waves),
the mechanism by which energy is transferred from the primary radiation beam to
biological targets is always via the secondary electrons which are produced. The
initial ionization events (which occur near-instantaneously at the microscopic level)
are the precursors to a chain of subsequent events which may eventually lead to the
clinical (macroscopic) manifestation of radiation damage.
Expression of cell death in individual lethally damaged cells occurs
later, usually at the point at which the cell next attempts to enter mitosis. Gross
(macroscopic and clinically observable) radiation effects are a result of the
wholesale functional impairment that follows from lethal damage being inflicted to
large numbers of cells or critical substructures. The timescale of the whole process
may extend to months or years. Thus, in clinical studies, any deleterious health
effects associated with a radiation procedure may not be seen until long after the
diagnostic test or treatment has been completed (Table2.1).
TABLE2.1. THE TIMESCALES OF RADIATION EFFECTS
Action
Approximate timescale
Initial ionizing event
1018 s
Transit of secondary electrons
1015 s
Production of ion radicals
1010 s
Production of free radicals
109 s
Chemical changes
105 s
Individual cell death

Gross biological effects
50
Hoursmonths
Hoursyears
BASIC RADIOBIOLOGY
2.3. BIOLOGICAL PROPERTIES OF IONIZING RADIATION

2.3.1. Types of ionizing radiation
In nuclear medicine, there are four types of radiation which play a relevant
role in tumour and normal tissue effects: gamma () radiation, beta () radiation,
alpha ()particles and Auger electrons.
2.3.1.1. Gamma radiation
Gamma radiation is an electromagnetic radiation of high energy
(usually above 25keV) and is produced by subatomic particle interactions.
Electromagnetic radiation is often considered to be made up of a stream of
wave-like particle bundles (photons) which move at the speed of light and whose
interaction properties are governed mainly by their associated wavelength.
Although the collective ionization behaviour of large numbers of photons can
be predicted with great accuracy, individual photon interactions occur at random
and, in passing through any type of matter, a photon may interact one or more
times, or never. In each interaction (which will normally involve a photoelectric
event, a Compton event or a pair production event), secondary particles are
produced, usually electrons (which are directly ionizing) or another photon
of reduced energy which itself can undergo further interactions. The electrons
undergo many ionizing events relatively close to the site of their creation and,
therefore, contribute mostly to the locally absorbed dose. Any secondary photons
which may be created carry energy further away from the initial interaction site
and, following subsequent electron-producing interactions, are responsible for
the dose deposition occurring at sites which are more distant from the original
interaction.
2.3.1.2. Beta radiation
Beta radiation is electrons emitted as a consequence of radionuclide decay.
A decay process can occur whenever there is a relative excess of neutrons ()
or protons (+). One of the excess neutrons is converted into a proton, with the
subsequent excess energy being released and shared between an emitted electron
and an anti-neutrino. Many radionuclides exhibit decay and, in all cases, the
emitted particle follows a spectrum of possible energies rather than being emitted
with a fixed, discrete energy. In general, the average energy is around one third
of the maximum energy. Most emitting radionuclides also emit photons as a
consequence of the initial decay, leaving the daughter nucleus in an excited,
metastable state. Since particles are electrons, once ejected from the host atom,
51
CHAPTER 2
they behave exactly as do the electrons created following the passage of a ray,
giving up their energy (usually of the order of several hundred kiloelectronvolts)
to other atoms and molecules through a series of collisions.
For radionuclides which emit both particles and photons, it is usually
the particulate radiation which delivers the greatest fraction of the radiation dose
to the organ which has taken up the activity. For example, about 90% of the dose
delivered to the thyroid gland by 131I arises from the component. On the other
hand, the emissions contribute more significantly to the overall whole body
dose.
2.3.1.3. Alpha particles
Alpha radiation is emitted when heavy, unstable nuclides undergo decay.
Alpha particles consist of a helium nucleus (two protons combined with two
neutrons) emitted in the process of nuclear decay. The particles possess
approximately 7000 times the mass of a particle and twice the electronic charge,
and give up their energy over a very short range (<100m). Alpha particles
usually possess energies in the megaelectronvolt range, and because they lose this
energy in such a short range are biologically very efficacious, i.e. they possess a
high linear energy transfer (LET; see Section 2.6.3) and are associated with high
relative biological effectiveness (RBE; see Section 2.6.4).
2.3.1.4. Auger electrons
Radionuclides which decay by electron capture or internal conversion
leave the atom in a highly excited state with a vacancy in one of the inner shell
electron orbitals. This vacancy is rapidly filled by either a fluorescent transition
(characteristic X ray) or non-radiative (Auger) transition, in which the energy
gained by the electron transition to the deeper orbital is used to eject another
electron from the same atom. Auger electrons are very short range, low energy
particles that are often emitted in cascades, a consequence of the inner shell atomic
vacancy that traverses up through the atom to the outermost orbital, ejecting
additional electrons at each step. This cluster of very low energy electrons can
produce ionization densities comparable to those produced by an particle track.
Thus, radionuclides which decay by electron capture and/or internal conversion
can exhibit high LET-like behaviour close (within 2 nm) to the site of the decay.
52
BASIC RADIOBIOLOGY
2.4. MOLECULAR EFFECTS OF RADIATION AND THEIR MODIFIERS

Radiation induced damage to biological targets may result from direct or
indirect action of radiation (Fig.2.1):
Direct action involves ionization or excitation (via Coulomb interactions) of
the atoms in the biological target. This gives rise to a chain of events which
eventually leads to the observable (macroscopic) damage. In normally
oxygenated mammalian cells, the direct effect accounts for about one third
of the damage for low LET radiations such as electrons and photons.
Indirect action involves radiation effects on atoms or molecules which
are not constituent parts of the biological target. Since cells exist in a rich
aqueous environment, the majority of indirect actions involve the ionization
or excitation of water molecules. The free radicals subsequently created
may then migrate and damage the adjacent biological targets. Indirect action
is the main cause of radiation damage and, in normoxic cells, accounts for
about two thirds of the damage.
Indirect action is predominant with low LET radiation, e.g. X and rays,
while direct action is predominant with high LET radiation, e.g. particles and
neutrons.
FIG.2.1. Illustration of the difference between direct and indirect damage to cellular DNA.
53
CHAPTER 2
2.4.1. Role of oxygen

Radiation effects may be influenced by several factors, especially the
presence or absence of oxygen. The free radicals (denoted by a dot placed to the
right of the atomic symbol) produced as a result of direct or indirect effects are
very reactive and seek to interact with other molecules which can share or donate
electrons. Molecular oxygen (O2) has two unpaired electrons and readily reacts
with free radicals, causing an increased likelihood that DNA (deoxyribonucleic
acid) will be damaged by the indirect process. Important reactions via which
oxygen can increase biological damage are:
R i + O 2 RO 2 i (highly toxic )
H i + O 2 HO 2 i
HO 2 i + HO 2 i = H 2O 2 (highly toxic) + O 2
where R represents an organic molecule.

The oxygen enhancement ratio (OER) is given by the dose in hypoxia (total
absence of oxygen) divided by the dose in air required to achieve an equivalent
biological effect. For low LET radiation, such as rays, the OER has a value
of ~3. For high LET radiation, such as particles, the OER decreases to almost
1.0.
2.4.2. Bystander effects
Bystander effects occur when a cell which has not been traversed by a
charged particle is damaged as a result of radiation interactions occurring in
neighbouring cells. The discovery of the bystander effect poses a challenge to
the traditional view that all radiation damage stems from direct interactions of
charged particles with critical cellular targets. For this reason, it still remains
controversial in radiobiology. A possible explanation is that irradiated cells may
send out a stress signal to nearby cells, which may elicit a response, e.g. the
initiation of apoptosis, in those cells. The overall relevance of the bystander
effect is presently difficult to gauge. It is probably most significant in radiation
protection considerations involving low doses since it amplifies the overall
radiation effect in situations where not all of the cells in a tissue are subjected
to particle transversal, i.e. the overall radiation risk to that tissue is higher than
would be expected from consideration of the gross response exhibited by those
cells which have been directly traversed by charged particles.
54
BASIC RADIOBIOLOGY
2.5. DNA DAMAGE AND REPAIR

2.5.1. DNA damage
DNA damage is the primary cause of cell death caused by radiation.
Radiation exposure produces a wide range of lesions in DNA such as single
strand breaks (SSBs), double strand breaks (DSBs), base damage, proteinDNA
cross-links and proteinprotein cross-links (seeFig.2.1). The number of DNA
lesions generated by irradiation is large, but there are a number of mechanisms
for DNA repair. As a result, the percentage of lesions causing cell death is very
small. The numbers of lesions induced in the DNA of a cell by a dose of 12 Gy
are approximately: base damages: >1000; SSBs: ~1000; DSBs: ~40. DSBs play
a critical role in cell killing, carcinogenesis and hereditary effects. There are
experimental data showing that the initially produced DSBs correlate with
radiosensitivity and survival at lower dose, and that unrepaired or misrepaired
DSBs also correlate with survival after higher doses. Furthermore, there is
experimental evidence for a causal link between the generation of DSBs and the
induction of chromosomal translocations with carcinogenic potential.
2.5.2. DNA repair
DNA repair mechanisms are important for the recovery of cells from
radiation and other damaging agents. There are multiple enzymatic mechanisms
for detecting and repairing radiation induced DNA damage. DNA repair
mechanisms, such as base excision repair, mismatch repair and nucleotide
excision repair, respond to damage such as base oxidation, alkylation and strand
intercalation. Excision repair consists of cleavage of the damaged DNA strand
by enzymes that cleave the polynucleotide chain on either side of the damage,
and enzymes which cleave the end of a polynucleotide chain allowing removal
of a short segment containing the damaged region. DNA polymerase can then
fill in the resulting gap using the opposite undamaged strand as a template. For
DSBs, there are two primary repair pathways, non-homologous end joining
(NHEJ) and homologous recombination. NHEJ repair operates on blunt ended
DNA fragments. This process involves the repair proteins recognizing lesion
termini, cleaning up the broken ends of the DNA molecule, and the final ligation
of the broken ends. DSB repair by homologous recombination utilizes sequence
homology with an undamaged copy of the broken region and, hence, can only
operate in late S/G2-phases of the cell cycle. Undamaged DNA from both strands
is used as a template to repair the damage. In contrast to NHEJ, the repair process
of homologous recombination is error-free. Repair by NHEJ operates throughout
the cell cycle but dominates in G1/S-phases. The process is error-prone because it
55
CHAPTER 2
does not rely on sequence homology. Unrepaired or misrepaired damage to DNA

will lead to mutations and/or chromosome damage in the exposed cell. Mutations
might lead to cancer or hereditary effects (when germ cells are exposed), whereas
severe chromosome damage often leads to cell death.
2.6. CELLULAR EFFECTS OF RADIATION
2.6.1. Concept of cell death
Radiation doses of the order of several grays may lead to cell loss. Cells
are generally regarded as having been killed by radiation if they have lost
reproductive integrity, even if they have physically survived. Loss of reproductive
integrity can occur by apoptosis, necrosis, mitotic catastrophe or by induced
senescence. Although all but the last of these mechanisms ultimately results in
physical loss of the cell, this may take a significant time to occur.
Apoptosis or programmed cell death can occur naturally or result from
insult to the cell environment. Apoptosis occurs in particular cell types after low
doses of irradiation, e.g. lymphocytes, serous salivary gland cells, and certain
cells in the stem cell zone in testis and intestinal crypts.
Necrosis is a form of cell death associated with loss of cellular membrane
activity. Cellular necrosis generally occurs after high radiation doses.
Reproductive cell death is a result of mitotic catastrophe (cells attempt
to divide without proper repair of DNA damage) which can occur in the first
few cell divisions after irradiation, and it occurs with increasing frequency after
increasing doses.
Ionizing radiation may also lead to senescence. Senescent cells are
metabolically active but have lost the ability to divide.
2.6.2. Cell survival curves
A quantitative understanding of many aspects of biological responses to
radiation may be made by consideration of the behaviour of the underlying cell
survival (dose response) characteristics. Although the practical determination of
cell survival curves is potentially fraught with experimental and interpretational
difficulties and is best performed by persons who are experts in such procedures,
an appreciation of the structure and meaning of such curves, even in a purely
schematic context, can be very helpful in understanding the role played by the
various factors which influence radiation response.
Figure2.2 shows the typical shape of a cell survival curve for mammalian
tissue. Physical radiation dose is plotted on the linear horizontal axis while
56
BASIC RADIOBIOLOGY
fractional cell survival is plotted on the logarithmic vertical axis. Each of the
individual points on the graph represents the fractional survival of cells resulting
from delivery of single acute doses of the specified radiation, which in this case
is assumed to be radiation. (In the context of the subject, an acute dose of
radiation may be taken to mean one which is delivered at high dose rate, i.e. the
radiation delivery is near instantaneous.) Mammalian cell survival curves plotted
in this way are associated with two main characteristics: a finite initial slope (at
zero dose) and a gradually increasing slope as dose is increased.
Surviving fraction
Dose (Gy)
FIG.2.2. A radiation cell survival curve plots the fraction of plated cells retaining colony
forming ability (cell surviving fraction) versus radiation absorbed dose.
2.6.3. Dose deposition characteristics: linear energy transfer

As noted above, the energy transfer to the absorbing medium (whether
that be animate or inanimate material) is via secondary electrons created by the
passage of the primary ionizing particle or ray. LET is a measure of the linear
rate at which radiation is absorbed in the absorbing medium by the secondary
particles and is defined by the International Commission on Radiation Units
and Measurements (ICRU) as being the quotient dE/dl, where dE is the average
energy locally imparted to the medium by a charged particle of specified energy
in traversing a distance dl. The unit usually employed for LET is kiloelectronvolt
per micrometre and some representative values are listed in Table2.2.
57
CHAPTER 2
TABLE
2.2. THE LINEAR ENERGY TRANSFER OF DIFFERENT
RADIATIONS
Radiation type
Linear energy transfer (keV/m)
60
0.2
250kVp X rays
2.0
10MeV protons
4.7
2.5MeV particles
166
1MeV electrons
0.25
10keV electrons
2.3
1keV electrons
12.3
Co rays
For radiobiological studies in particular, the concept of LET is problematic

since it relates to an average linear rate of energy deposition but, at the
microscopic level (i.e. at dimensions comparable with the critical cellular targets),
the energy deposited per unit length along different parts of a single track may
vary dramatically. In particular, as charged particles lose energy in their passage
through a medium via the result of collision and ionizing processes, the LET rises
steeply to its highest value towards the very end of their range. The change in
LET value along the track length is one reason why average LET values correlate
poorly with observed (i.e. macroscopic) biological effects. For these reasons, the
directly measured RBE is of much greater use as an indicator of the differing
biological efficacies of various radiation types.
2.6.4. Determination of relative biological effectiveness
For a given biological end point, the RBE of the high LET radiation is
defined as the ratio of the isoeffective doses for the reference (low LET) and the
high LET radiation (Fig.2.3). The reference radiation is usually 60Co g rays or
high energy (250kVp) X rays.
58
Surviving fraction
BASIC RADIOBIOLOGY
Dose (Gy)
FIG.2.3. The relative biological effectiveness of a radiation is defined as the ratio of the
dose required to produce the same reduction in cell survival as a reference low linear energy
transfer (LET) radiation.
If the respective low and high LET isoeffective doses are dL and dH, then:
RBE =
dL
(2.1)
dH
If the basic cell survival curves are described in terms of the

linearquadratic (LQ) model, then the surviving fraction S as a function of acute
doses at low and high LET is respectively given as:
S L = exp( L d L L d L2 ) (2.2)
2
S H = exp( H d H H d H
) (2.3)
where the suffixes L and H again respectively refer to the low and high LET
instances.
Figure2.4 shows an example of how the RBEs determined at any particular
end point (cell surviving fraction) vary with changing dose for a given radiation
fraction size for a low LET radiation. The maximum RBE (RBEmax) occurs
at zero dose and, in terms of microdosimetric theory, corresponds to the ratio
59
CHAPTER 2
between the respective high and low LET linear radiosensitivity constants, H
and L, i.e.:
H

L
(2.4)
RBE
RBE max =
Dose/fraction
FIG.2.4. Relative biological effectiveness (RBE) as a function of the radiation dose per
fraction.
If the quadratic radiosensitivity coefficients (H and L) are unchanged with

changing LET (i.e. H=L), then, at high doses, the RBE tends to unity. However,
this constancy of , assumed by the theory of Kellerer and Rossi, has been
challenged and, if does change with LET, then RBE will tend asymptotically to
an alternativeminimum value (RBEmin) given by:
RBE min =
H
(2.5)
L
and the working RBE at any given dose per fraction is given as:
2
2
+ 4d L RBE min
( / )L RBE max + ( / )L RBE max
( / )L + d L (2.6)
RBE =
2 ( / )L + d L
60
BASIC RADIOBIOLOGY
when expressed in terms of the low LET dose per fraction dL or:
RBE =
2
2
( / )L + ( / )L + 4d H ( / )L RBE max + RBE min
d H
2d H
(2.7)
when expressed in terms of the high LET dose per fraction dH.
Figure2.4 was derived using RBEmax =5, RBEmin=1 and (/)L=3 Gy,
but the general trend of a steadily falling RBE with increasing dose per fraction
is independent of the chosen values. Clearly, the assumption of a fixed value
of RBE, if applied to all fraction sizes, could lead to gross clinical errors and
Eqs(2.6) and (2.7) make the point that determination of RBEs in a clinical setting
is potentially complex and will depend on accurate knowledge of RBEmax and (if
it is not unity) RBEmin. Although there is not yet clear evidence over whether
or not there is a consistent trend for RBEmin to be non-unity, the possibility is
nevertheless important as it may hold very significant implications.
Figure2.4 also shows schematically how the rate of change of RBE with
changing dose per fraction is influenced by the existence of a non-unity RBEmin
parameter. Even for a fixed value of RBEmax, the potential uncertainty in the RBE
values at the fraction sizes likely to be used clinically might themselves be very
large if RBEmin is erroneously assumed to be unity. These uncertainties would be
compounded if there were an additional linkage between RBEmax and the tissue
/ value.
As is seen from Eqs(2.6) and (2.7), the RBE value at any particular dose
fraction size will also be governed by the low LET / ratio (a tissue dependent
parameter which provides a measure of how tissues respond to changes in dose
fractionation) and the dose fraction size (a purely physical parameter) at the point
under consideration. Finally, and as has been shown through the earlier clinical
experience with neutron therapy, the RBEmax value may itself be tissue dependent,
likely being higher for the dose-limiting normal tissues than for tumours. This
tendency is borne out by experimental evidence using a variety of ion species as
well as by theoretical microdosimetric studies. This potentially deleterious effect
may be offset by the fact that, in carbon-, helium- and argon-ion beams, LET
(and, hence, RBE) will vary along the track in such a way that it is low at the
entry point (adjacent to normal tissues) and highest at the Bragg peak located
in the tumour. However, although this might be beneficial, it does mean that the
local RBE is more spatially variable than is indicated by Eq.(2.6).
Owing to the difficulties in setting reference doses at which clinical
inter-comparisons could be made more straightforward, Wambersie proposed
that a distinction be made between the reference RBE and the clinical
61
CHAPTER 2
RBE. Thus, the reference RBE might be that determined at 2 Gy fractions on

a biological system end point representative, for example, of the overall late
tolerance of normal tissues. As more clinical experience of using the particular
radiation becomes available, a more practical clinical RBE evolves, this being
the reference RBE empirically weighted by collective clinical experience and
by volume effects related to the beam characteristics, geometry or technical
conditions.
2.6.5. The dose rate effect and the concept of repeat treatments
When mammalian cells are irradiated, it is helpful to visualize their
subsequent death as resulting from either of two possible processes. In the first
process, the critical nuclear target (DNA) is subjected to a large deposition of
energy which physically breaks both strands of the double helix structure and
disrupts the code sufficiently to disallow any opportunity of repair. This process
can be thought of as a single-hit process and the total amount of DNA damage
created this way is directly proportional to the dose delivered.
In the second process, an ionizing event occurs and releases only sufficient
energy to disrupt the coding carried by one strand of the DNA. Following this
event, and if the irradiation continues, two outcomes are possible: either the
broken strand will restore itself to its original state (no lethality) or, prior to full
repair taking place, a second, independent radiation event may occur in the same
location and damage the opposite strand of the DNA, a complementary action
between the two damaged strands then leading to cell lethality in what is called a
two-hit process. Since this route depends on there being two independent events,
each having a probability proportional to dose, the number of damaged DNA
targets created this way is proportional to dose dose, i.e. dose2. Once created,
the radiation damage due to these two possible routes is indistinguishable
(i.e. both processes are lethal). From this simplified description, it is clear that the
observed radiation response characterized in the cell survival curve will consist
of two components: one linear with dose and the other quadratic, i.e. proportional
to dose2. This phenomenological description qualitatively explains the shape of
a radiation survival curve, with a finite initial slope at low dose followed by an
increasingly downward curvature as dose increases.
However, the amount of damage created in the second process is dependent
on the ability of the second break to be induced before the first break has repaired
itself and, thus, is dependent on the dose rate.
Figure2.5 shows a range of response curves in which the doses are
delivered at four different dose rates, the individual doses taking proportionately
longer to deliver as dose rate is reduced. This graph illustrates that reducing the
dose rate causes the overall shape of the response curve to become less curvy
62
BASIC RADIOBIOLOGY
than in the acute case, but that the initial slope remains unchanged. When the
doses are all delivered at a very low dose rate, as is the case for most radionuclide
therapies, the response is essentially a straight line, when the curves are plotted
on a log-linear scale, as is common practice for radiation survival curves. This
means that the low dose response is purely exponential.
Surviving fraction
Dose (Gy)
FIG.2.5. Surviving fraction as a function of dose for different dose rates. It is important
to note that most radionuclide therapies are delivered at low dose rate in the range of
0.10.5 Gy/h, when the survival curve is almost linear.
2.6.6. The basic linearquadratic model

The basic equation describing the shape of the cell survival curves shown
in Fig.2.2 is referred to as the LQ model, which has a biophysical origin. Cell
survival following delivery of an acute dose d is given as:
S = exp( d d 2 ) (2.8)
where (in units of Gy1) and (in units of Gy2) are the respective linear and
quadratic sensitivity coefficients.
63
CHAPTER 2
If the treatment is repeated in N well spaced fractions, then the net survival
is SN, where:
S N = S N = exp(N d N d 2 ) (2.9)
Taking natural logarithms on both sides of Eq.(2.9) and dividing throughout

by leads to:
ln S N
= Nd
Nd 2
(2.10)
( / )
2.6.7. Modification to the linearquadratic model

for radionuclide therapies
Targeted radionuclide therapy normally involves irradiation of the
tumour/normal tissues at a dose rate which is not constant but which reduces as
treatment proceeds, as a consequence of the combination of radionuclide decay
and biological clearance of the radiopharmaceutical. To allow for this, a more
extensive formulation of the LQ model is required.
2.6.8. Quantitative intercomparison of different treatment types
In many aspects of LQ modelling, a term called the biological effective
dose (BED) is employed to assess and inter-compare different treatment types.
BED is defined as:
BED =
ln S N
d
(2.11)
= Nd 1 +
( / )
Although the parameters and are rarely known in detail for individual
tumours or tissues, values of the ratio / (in units of grays) are becoming
increasingly known from clinical and experimental data. In general, / is
systematically higher (520 Gy) for tumours than for critical, late-responding
normal tissues (25 Gy) and it is this difference which provides the BED concept
with much of its practical usefulness.
64
BASIC RADIOBIOLOGY
For non-acute treatments (those in which the dose delivery is protracted

over a long time period on account of a lower dose rate), the BED is re-written
as:
d g (t )
(2.12)
BED = Nd 1 +
( / )
where g(t) is a function of the time t taken for delivery:

g (t ) =
2 1 exp( t )
1
(2.13)
and where is the mono-exponential time constant relating to the repair of

sublethal damage. is related to the tissue repair half-time (T1/2) via:
=
0.693
(2.14)
T1/2
For a treatment delivery at constant dose rate R, the delivered dose d is

related to treatment time t via d=Rt, thus:
2 R 1 exp( t )
BED = Rt 1 +
1
(2.15)
( / )
t
When t > 12 h, Eq.(12.15) simplifies to:
2 R
BED = Rt 1 +
(2.16)
( / )
2.6.9. Cellular recovery processes

At lower doses and dose rates, cellular recovery may play an important role
in the fixation of the radiation damage. There are three broad types of cellular
radiation damage:
(a) Lethal damage in which the cellular DNA is irreversibly damaged to such
an extent that the cell dies or loses its proliferative capacity;
(b) Sublethal damage in which partially damaged DNA is left with sufficient
capacity to restore itself over a period of a few hours, provided there is no
further damage during the repair period;
65
CHAPTER 2
(c) Potentially lethal damage in which repair of what would normally be

a lethal event is made possible by manipulation of the post-irradiation
cellular environment.
2.6.10. Consequence of radionuclide heterogeneity
The effectiveness per unit dose of a radiopharmaceutical depends on the
heterogeneity of the radionuclide distribution. Global non-uniformity of a source
distribution, which results in pockets of cells (tumour or normal tissue) receiving
less than the average dose will almost always result in a greater fraction of cell
survivors, than if all cells receive a uniform dose. The one possible exception
would be if a radiopharmaceutical would selectively localize at sensitive target
cells, within an organ, that are key for organ regeneration or function, e.g. crypt
cells in the colon. The cellular response also depends on the microdosimetry,
especially if the radiopharmaceutical in question selectively localizes on the cell
surface or internalizes within a certain cohort of cells within a tumour/normal
organ. Radiolabels that selectively localize on the surface of cells or are
internalized may exhibit geometric enhancement factors that modulate a response.
The reader is referred to ICRU Report 67 on absorbed dose specification in
nuclear medicine for more details.
2.7. GROSS RADIATION EFFECTS ON TUMOURS
AND TISSUES/ORGANS
2.7.1. Classification of radiation damage (early versus late)
Cells which are lethally affected by radiation may continue to function
for some time after the infliction of the damage, only dying when attempting
to undergo subsequent cell division (mitosis). Clinically observed radiation
effects in whole tissues or organs reflect the damage inflicted to large numbers
of constituent cells and, thus, appear on a timescale which is governed largely
by the underlying proliferation rates of those cells. Such observable effects
are classified as being either late or early, depending on the speed at which
they manifest themselves following irradiation. Late effects appear months or
years after irradiation and appear in structures which proliferate very slowly,
e.g. kidney. Early (or acute) effects appear within days, weeks or months of
irradiation and are associated with fast-proliferating epithelial tissues, e.g. bone
marrow, mucosa, intestinal tract, etc.
66
BASIC RADIOBIOLOGY
In most types of radiotherapy, it is the late effects which are considered

to be most critical and which generally limit the total dose which may be
delivered to the tumour. If the radiation tolerance of the late-responding tissues is
exceeded, then the subsequent reactions, depending on the tissues in which they
arise, may seriously affect mobility and/or quality of life, and may even be life
threatening. Such problems arise long after the completion of treatment and are,
thus, impossible to correct. These are the serious considerations which are at the
heart of the therapeutic index concept discussed below (seeSection 2.7.3). Acute
reactions in radiotherapy, although they may be unpleasant, are usually transient
and easier to control by adjustment of the treatment dose delivery pattern and/or
simple medication. In radionuclide therapies, it is in most instances possible
to circumvent acute radiation toxicities once they begin to occur, such as by
accelerating clearance of the radiopharmaceutical. Chronic toxicities, such as
to the kidney, usually occur at times which are long relative to the lifetime of
the radionuclide. Hence, considerable importance should be attributed to the
administration of safe activities of therapeutic radionuclides that do not exceed
any dose limiting constraints.
2.7.2. Determinants of tumour response
Irrespective of the mechanism used to achieve tumour targeting, the
potential advantage of radionuclide therapy over other forms of radiation therapy
is its ability to deliver dose to both the local disease and to occult tumour deposits.
In nuclear medicine, the primary determinants of treatment effectiveness
are:
The tumour specificity of the radionuclide carrier.
The homogeneity of uptake of the carrier within the targeted tumour(s).
The intrinsic RBE (seeSection 2.6.4) of the radiation used for the therapy:
this is determined primarily by the nature of the radionuclide emissions
(e.g. particles, particles, low energy rays, Auger electrons, etc.).
The range of the particles, as determined by their energies.
The total dose delivered.
The responsiveness of the targeted tumour cells to radiation. This will be
determined by radiobiological properties such as cellular radiosensitivity
and the variations of sensitivity within the cell cycle, the oxygen status of
the cells (fully oxic, partially oxic or hypoxic), the ability of the cells to
recover from sublethal radiation damage and the degree to which tumour
growth (repopulation) may occur during therapy.
67
CHAPTER 2
These factors are complementary and interactive, and should not

be considered in isolation from each other. Thus, for example, significant
non-uniformity of uptake within the tumour may result in dose cold spots, but
the detrimental potential of these might be offset by the selection of a radionuclide
which emits particles of sufficient range to produce a cross-fire effect within the
cold spots from those adjacent cells which are properly targeted. The significance
of cold spot and cross-fire effects is further dependent on the size of the tumour
deposit under consideration.
2.7.3. The concept of therapeutic index in radiation
therapy and radionuclide therapy
The therapeutic index of a particular radiation treatment (often referred to in
older publications as the therapeutic ratio) is a measure of the resultant damage
to the tumour vis a vis the damage to critical normal structures. Treatments with
a high therapeutic index will demonstrate good tumour control and low normal
tissue morbidity; treatments with a low therapeutic index will be associated with
a low tumour control and/or high morbidity. There have been several attempts to
provide quantitative definitions of therapeutic index, but it is usually sufficient to
consider therapeutic index as being a qualitative concept any new treatment
which, relative to an existing treatment, improves tumour control and/or reduces
morbidity is said to be associated with an improved therapeutic index.
In conventional (external beam) radiotherapy, the normal tissues at risk
will be those immediately adjacent to the tumour being treated. Doses to the
normal tissues (along with the risk of toxicity) may be reduced by attention to
a combination of physical and radiobiological factors. In targeted radionuclide
therapy, the tumour may be single and discrete (as is the case in most external
beam therapy) or may consist of distributed masses or metastatic deposits at
several locations within the body. The normal tissues at risk may themselves be
widely distributed but, more particularly, may be a reflection of the particular
uptake pattern of the targeting compound being used for the therapy.
2.7.4. Long term concerns: stochastic and deterministic effects
The radiation detriment which results from radiation exposure may be
classified as being either stochastic or deterministic in nature. Stochastic effects
(e.g. hereditary damage, cancer induction) are those for which the likelihood
of them occurring is dose related, but the severity of the resultant condition is
not related to the dose received. Deterministic effects (e.g. cataract induction,
general radiation syndromes, bone marrow ablation, etc.) manifest themselves
with a severity which is dose related. In general, it is predominantly stochastic
68
BASIC RADIOBIOLOGY
effects which need to be considered as potential side effects from diagnostic uses
of radionuclides, although deterministic damage may result if the embryo or fetus
is irradiated. For radionuclide therapy applications, the concerns relate to both
stochastic and deterministic effects.
2.8. SPECIAL RADIOBIOLOGICAL CONSIDERATIONS
IN TARGETED RADIONUCLIDE THERAPY
2.8.1. Radionuclide targeting
Tumour targeted radiotherapy is a very promising approach for the
treatment of wide-spread metastasis and disseminated tumour cells. This
technique aims to deliver therapeutic irradiation doses to the tumour while
sparing normal tissues by targeting a structure that is abundant in tumour cells,
but rare in normal tissues. This can be done by using antibodies labelled with
a therapeutic relevant radionuclide acting against a specific tumour target.
Radiolabelled antibody therapy has already become common in the treatment of
non-Hodgkins lymphoma, e.g. 131I-tositumomab (Bexxar) and 90Y-ibritumomab
tiuxetan (Zevalin), and exhibits great potential for being extended to other
diseases. A good example is epidermal growth factor (EGF) labelled with 125I
which will bind EGF receptors. EGF receptors are overexpressed on tumour cells
in many malignancies such as highly malignant gliomas. At present, several other
radiolabelled antibodies are being used in experimental models and in clinical
trials to study their feasibility in other types of cancer.
2.8.2. Whole body irradiation
Conventional external beam radiotherapy involves controlled irradiation of
a carefully delineated target volume. Normal structures adjacent to the tumour
will likely receive a dose, in some cases a moderately high dose, but the volumes
involved are relatively small. The rest of the body receives only aminimal dose,
mostly arising from radiation scattered within the patient from the target volume
and from a small amount of leakage radiation emanating from the treatment
machine outside the body.
Targeted radionuclide therapies are most commonly administered
intravenously and, thus, can give rise to substantial whole body doses and, in
particular, doses to the radiation sensitive bone marrow. Once the untargeted
activity is removed from the blood, it may give rise to substantial doses in
normal structures, especially the kidneys. Furthermore, the activity taken up by
69
CHAPTER 2
the kidneys and targeted tumour deposits may (if ray emissions are involved)
continue to irradiate the rest of the body.
2.8.3. Critical normal tissues for radiation and radionuclide therapies
Since the radiation doses used in radionuclide therapies are much higher
than the doses used for diagnosis, (prolonged) retention of the pharmaceuticals
within the blood circulation and, hence, increased accumulation of radionuclides
in non-tumour cells, might lead to unwanted toxicities. The bone marrow, kidney
and liver are regarded as the main critical organs for systemic radionuclide therapy.
Other organs at risk are the intestinal tract and the lungs. The bone marrow is
very sensitive towards ionizing radiation. Exposure of the bone marrow with high
doses of radiation will lead to a rapid depression of white blood cells followed
a few weeks later by platelet depression, and in a later stage (approximately one
month after exposure) also by depression of the red blood cells. In general, these
patients could suffer from infections, bleeding and anaemia. Radiation damage to
the gastrointestinal tract is characterized by de-population of the intestinal mucosa
(usually between 3 and 10 days) leading to prolonged diarrhoea, dehydration,
loss of weight, etc. The kidneys, liver and lungs will show radiation induced
damage several months after exposure. In kidneys, a reduction of proximal tubule
cells is observed. These pathological changes finally lead to nephropathy. In the
liver, hepatocytes are the radiosensitive targets. Since the lifespan of the cells is
about a year, deterioration of liver function will become apparent between 3 and
9 months after exposure. In lungs, pulmonary damage is observed in two waves:
an acute wave of pneumonitis and later fibrosis.
The determinants of normal tissue response from radionuclide studies
is a large subject due to the diversity of radiopharmaceuticals with differing
pharmacokinetics and biodistribution, and the widely differing responses and
tolerances of the critical normal tissues. A principal determinant of the type of
toxicity depends on the radionuclide employed. For example, isotopes of iodine
localize in the thyroid (unless blocked), salivary glands, stomach and bladder.
Strontium, yttrium, samarium, fluorine, radium, etc. concentrate in bone.
Several radiometals, such as bismuth, can accumulate in the kidney. If these
radionuclides are tightly conjugated to a targeting molecule, the biodistribution
and clearance are determined by that molecule. For high molecular weight
targeting agents, such as an antibody injected intravenously, the slow plasma
clearance results in marrow toxicity being the principal dose limiting organ. For
smaller radiolabelled peptides, renal toxicity becomes of concern. When studying
a new radiopharmaceutical or molecular imaging agent, it is always important
to perform a detailed study of the biodistribution at trace doses, to ensure the
70
BASIC RADIOBIOLOGY
absence of radionuclide sequestration within potentially sensitive tissue, such as

the retina of the eye or the germ cells of the testes.
A review of normal tissue toxicities resulting from radionuclide therapies is
given by Meredith et al. (2008).
2.8.4. Imaging the radiobiology of tumours
The development of molecular imaging using positron emission tomography
(PET) has given rise to new radiotracers which have the potential to assess
several features of radiobiological relevance for therapy planning. One tracer
that is becoming more widely available for PET imaging is fluorothymidine.
This radiotracer exhibits the property of becoming selectively entrapped within
cells that are progressing through S-phase (DNA replication) of the cell cycle,
thus providing a signal which should be proportional to cell proliferation,
andminimizing the signal from cells in G0 or in cell cycle arrest. The ability to
selectively identify only replicating cells separate from all tumour cells present
within the computed tomography-determined tumour volume may present an
excellent opportunity for more accurate measures of the initial viable tumour
burden as well as evaluating tumour response. Complementary to measuring
tumour response is the measurement of therapeutic efficacy through radiotracers
that selectively target cell death. Radiotracers are under development with the
ability to selectively bind to receptors expressed on cells undergoing programmed
cell death, e.g. radiolabelled annexin V. Another area of active research is in the
field of hypoxia imaging. Cells within a tumour microenvironmental region of low
partial oxygen pressure, i.e. hypoxia, are known to exhibit a great radio-resistance
to both radiation and chemotherapy relative to those under normoxic conditions.
A number of PET radiotracers are under evaluation for imaging tumour
hypoxia with PET, including fluoromisonidazole (18F-FMISO), fluoroazomycin
arabinoside (18F-FAZA) and copper-diacetyl-bis(N4-methylthiosemicarbazone)
(64Cu-ATSM). The ability to measure the radiobiological attributes of a tumour
prior to therapy may provide invaluable information concerning the relative
resistance/aggressiveness of tumours, leading to improved management of these
patients.
2.8.5. Choice of radionuclide to maximize therapeutic index
The choice of the optimum radionuclide to maximize the therapeutic index
depends on a number of factors. First, the range of the emitted particles from the
radionuclide should depend on the type of tumour being treated. For leukaemia
or micrometastatic deposits, consisting of individual or small clusters of tumour
cells, there is a distinct advantage of using radionuclides which emit very short
71
CHAPTER 2
range particles. Since particles have ranges of <100 m in tissue, particle

emitters would have an advantage, if the targeting molecule were able to reach
all tumour cells. However, particle emitting radionuclides are not widely
available and are extremely expensive. In addition, the short range of particles
can be a disadvantage for bulk tumours. For these reasons, almost all therapeutic
radionuclides utilized in the clinic today consist of medium (131I) or long range
(90Y, 186Re) emitters. These radionuclides are advantageous when treating solid
tumours for which target receptor (antigen) expression may be heterogeneous, or
with non-uniform delivery, due to the greater cross-fire range of their emissions
(ranging up to a 1cm range in unit density tissue).
A second important consideration is the choice of radionuclide half-life.
If the half-life is too short, then the radiolabelled tumour targeting agent may
have insufficient time to reach its target, resulting in aminimal therapeutic index.
Increasing the half-life will increase the therapeutic index, but render the patient
radioactive for a longer period of time, resulting in prolonged confinement, greater
expense and radiation risks to staff and family. Pure emitting radionuclides such
as 90Y and 32P have advantages in that theyminimize the exposure to personnel
assisting the patient. The half-life of the radionuclide should ideally match the
biological uptake and retention kinetics of the tumour-targeting carrier used. For
large protein carriers such as antibodies, radionuclides with half-lives of several
days are required to optimize the therapeutic index. For smaller molecular
targeting agents such as peptides, short lived radionuclides may be better suited
tominimize radioactive waste.
Thirdly, it is necessary to consider radiochemistry, ease and stability of the
radiolabelled end product. All of these factors need to be taken into consideration
in order to produce the optimum therapeutic targeting compound for use in
clinical therapeutic applications.
BIBLIOGRAPHY
DALE, R.G., JONES, B. (Eds), Radiobiological Modelling in Radiation Oncology, The British
Institute of Radiology, London (2007).
HALL, E.J., GIACCIA, A.J., Radiobiology for the Radiologist, 6th edn, Lippincott, Williams
and Wilkins, Philadelphia, PA (2006).
INTERNATIONAL COMMISSION ON RADIATION UNITS, Absorbed-dose Specification
in Nuclear Medicine, Rep. 67, Nuclear Technology Publishing, Ashford, United Kingdom
(2002).
MEREDITH, R., WESSELS, B., KNOX, S., Risks to normal tissue from radionuclide therapy,
Semin. Nucl. Med. 38 (2008) 347357.
72
CHAPTER 3
RADIATION PROTECTION
S.T. CARLSSON
Department of Diagnostic Radiology,
Uddevalla Hospital,
Uddevalla, Sweden
J.C. LE HERON
Division of Radiation, Transport and Waste Safety,
Vienna
3.1. INTRODUCTION
Medical exposure is the largest human-made source of radiation exposure,
accounting for more than 95% of radiation exposure. Furthermore, the use of
radiation in medicine continues to increase worldwide more machines are
accessible to more people, the continual development of new technologies and
new techniques adds to the range of procedures available in the practice of
medicine, and the role of imaging is becoming increasingly important in day
to day clinical practice. The introduction of hybrid imaging technologies, such
as positron emission tomography/computed tomography (PET/CT) and single
photon emission computed tomography (SPECT)/CT, means that the boundaries
between traditional nuclear medicine procedures and X ray technologies are
becoming blurred. Worldwide, the total number of nuclear medicine examinations
is estimated to be about 35 million per year.
In Chapter2, basic radiation biology and radiation effects were described,
demonstrating the need for a system of radiation protection. Such a system allows
the many beneficial uses of radiation to be utilized, but at the same time ensures
detrimental radiation effects are either prevented orminimized. This can be
achieved by having the objectives of preventing the occurrence of deterministic
effects and of limiting the probability of the stochastic effects to a level that is
considered acceptable. In a nuclear medicine facility, consideration needs to
be given to the patient, the staff involved in performing the nuclear medicine
procedures, members of the public and other staff that may be in the nuclear
medicine facility, carers and comforters of patients undergoing procedures,
73
CHAPTER 3
and persons who may be undergoing a nuclear medicine procedure as part of a

biomedical research project.
This chapter discusses how the objectives stated above are achieved
through a system of radiation protection, and how such a system should be
applied practically in a hospital in general and in nuclear medicine specifically.
3.2. BASIC PRINCIPLES OF RADIATION PROTECTION
The means for achieving the objectives of radiation protection have evolved
over many years to the point where, for some time, there has been a reasonably
consistent approach throughout the world namely the system of radiological
protection, as espoused by the International Commission on Radiological
Protection (ICRP). The following will briefly describe this system, specifically as
it applies to nuclear medicine.
3.2.1. The International Commission on Radiological
Protection system of radiological protection
The principles of radiation protection and safety upon which the IAEA
safety standards are based are those developed by the ICRP. The detailed
formulation of these principles can be found in ICRP publications and they
cannot easily be paraphrased without losing their essence. However, a brief,
although simplified, summary of the principles is given in this section.
The ICRP recommends a system of radiological protection to cover all
possible exposure situations. There are many terms associated with the system
and some of these will now be introduced.
The ICRP in its Publication 103 [3.1] divides all possible situations
of where exposure can occur into three types planned exposure situations,
emergency exposure situations and existing exposure situations. For the practice
of nuclear medicine, only the first situation is relevant. The use of radiation
in nuclear medicine is a planned exposure situation it needs to be under
regulatory control, with an appropriate authorization in place from the regulatory
body before operation can commence. Misadministration, spills and other such
incidents or accidents can give rise to what is called potential exposure, but these
remain part of the planned exposure situation as their occurrence is considered
in the granting of an authorization. It should be noted that the ICRP has used the
term practice to describe a planned exposure situation such as the operation of a
nuclear medicine facility.
74
The ICRP then puts exposure of individuals into three categories medical
exposure, occupational exposure and public exposure:
Medical exposure refers primarily to exposure incurred by patients for
the purpose of medical diagnosis or treatment. It also refers to exposures
incurred by individuals helping in the support and comfort of patients
undergoing diagnosis or treatment, and by volunteers in a programme of
biomedical research involving their exposure.
Occupational exposure is the exposure of workers incurred in the course of
their work.
Public exposure is exposure incurred by members of the public from all
exposure situations, but excluding any occupational or medical exposure.
All three need to be considered in the nuclear medicine facility.
An individual person may be subject to one or more of these categories of
exposure, and for radiation protection purposes such exposures are dealt with
separately.
The ICRP system has three fundamental principles of radiological
protection, namely:
The principle of justification: Any decision that alters the radiation exposure
situationshould do more good than harm.
The principle of optimization of protection: The likelihood of incurring
exposures, the number of people exposed and the magnitude of their
individual doses should all be kept as low as reasonably achievable
(ALARA), taking into account economic and societal factors.
The principle of limitation of doses: The total dose to any individual
from regulated sources in planned exposure situations other than medical
exposure of patients should not exceed the appropriate limits recommended
by the ICRP. Recommended dose limits are given in Table3.1.
In a nuclear medicine facility, occupational and public exposures are subject
to all three principles, whereas medical exposure is subject to the first two only.
More detail on the application of the ICRP system for radiological protection as
it applies to a nuclear medicine facility is given in the remainder of this chapter.
75
CHAPTER 3
TABLE3.1. RECOMMENDED DOSE LIMITS IN PLANNED EXPOSURE

SITUATIONSa
Type of limit
Effective dose
Annual equivalent dose in:
Lens of the eyed
Skine, f
Hands and feet
a
Occupational
Public
20 mSv per year, averaged over defined

periods of 5 yearsb
1 mSv in a yearc
20 mSv
500 mSv
500 mSv
15 mSv
50 mSv
Limits on effective dose are for the sum of the relevant effective doses from external
exposure in the specified time period and the committed effective dose from intakes of
radionuclides in the same period. For adults, the committed effective dose is computed for a
50 year period after intake, whereas for children it is computed for the period up to reaching
70 years of age.
With the further provision that the effective dose should not exceed 50 mSv in any single
year. Additional restrictions apply to the occupational exposure of pregnant women.
In special circumstances, a higher value of effective dose could be allowed in a single year,
provided that the average over 5 years does not exceed 1 mSv/a.
In 2011, the ICRP recommended that the occupational dose limit be lowered from the
previous 150 mSv/a to 20 mSv/a, averaged over 5 years, and with no more than 50 mSv in
any single year.
The limitation on effective dose provides sufficient protection for the skin against stochastic
effects.
Averaged over a 1cm2 area of skin regardless of the area exposed.
3.2.2. Safety standards

Safety standards are based on knowledge of radiation effects and on the
principles of protection described above. In this respect, the development of
safety standards by the IAEA follows a well established approach. The United
Nations Scientific Committee on the Effects of Atomic Radiation (UNSCEAR),
a body set up by the United Nations in 1955, compiles, assesses and disseminates
information on the health effects of radiation and on levels of radiation exposure
due to different sources; this information was taken into account in developing
the standards. Following a decision made in 1960, the IAEA safety standards
are based on the recommendations of the ICRP, which also take account of the
scientific information provided by UNSCEAR.
Purely scientific considerations, however, are only part of the basis for
decisions on protection and safety, and the safety standards implicitly encourage
76
decision makers to make value judgements about the relative importance of

different kinds of risks and about the balancing of risks and benefits. General
acceptance of risk is a matter of consensus and, therefore, international safety
standards should provide a desirable international consensus for the purpose of
protection.
For these reasons, international consensus is integral to the IAEA safety
standards, which are prepared with the wide participation of and approval by its
Member States and relevant international organizations. The current version of
what is commonly called the Basic Safety Standards (BSS) is entitled Radiation
Protection and Safety of Radiation Sources: International Basic Safety Standards
(2014) [3.2]. The BSS are jointly sponsored by the European Commission,
the Food and Agriculture Organization of the United Nations, the IAEA, the
International Labour Organization, the OECD Nuclear Energy Agency, the
Pan American Health Organization (PAHO), the United Nations Environment
Programme and the World Health Organization (WHO).
The BSS comprises five sections: Introduction, General requirements for
protection and safety, Planned exposure situations, Emergency exposure situations
and Existing exposure situations, as well as four schedules. The purpose of the
BSS is to establish basic requirements for protection against exposure to ionizing
radiation and for the safety of radiation sources that may deliver such exposure.
The requirements of the BSS underpin the implementation of radiation protection
in a nuclear medicine facility, supplemented by the relevant IAEA Safety Guides
and Safety Reports.
3.2.3. Radiation protection quantities and units
The basic dosimetry quantity for use in radiation protection is the mean
organ or tissue dose DT given by:
DT = T / mT
(3.1)
where
mT is the mass of the organ or tissue T;
and T is the total energy imparted by radiation to that tissue or organ.
The International System of Units (SI) unit of mean organ dose is joules per
kilogram (J/kg) which is termed gray (Gy).
Owing to the fact that different types of ionizing radiation will have
different effectiveness in damaging human tissue at the same dose, and the fact
77
CHAPTER 3
that the probability of stochastic effects will depend on the tissue irradiated, it is
necessary to introduce quantities to account for these factors. Those quantities
are equivalent dose and effective dose. Since they are not directly measurable,
the International Commission on Radiation Units and Measurements (ICRU)
has defined a set of operational quantities for radiation protection purposes (area
monitoring and personal monitoring): the ambient dose equivalent, directional
dose equivalent and personal dose equivalent.
Regarding internal exposure from radionuclides, the equivalent dose
and the effective dose are not only dependent on the physical properties of the
radiation but also on the biological turnover and retention of the radionuclide.
This is taken into account in the committed dose quantities (equivalent and
effective).
3.2.3.1. Equivalent dose
It is a well known fact in radiobiology that densely ionizing radiation such
as particles and neutrons will cause greater harm to a tissue or organ than
rays and electrons at the same mean absorbed dose. This is because the dense
ionization events will result in a higher probability of irreversible damage to the
chromosomes and a lower chance of tissue repair. To account for this, the organ
dose is multiplied with a radiation weighting factor in order to get a quantity that
more closely reflects the biological effect on the irradiated tissue or organ. This
quantity is called the equivalent dose and is defined as:
H T = wr DT,r (3.2)
where
DT,R is the mean tissue or organ dose delivered by type R radiation;
and wR is the radiation weighting factor.
For X rays, rays and electrons, wR =1; for particles, wR =20. The SI unit of
equivalent dose is joules per kilogram (J/kg), which is termed sievert (Sv). In a
situation of exposure from different types of radiation, the total equivalent dose is
the sum of the equivalent dose from each type of radiation.
78
3.2.3.2. Effective dose

The relationship between the probability of stochastic effects and equivalent
dose is found to depend on the organ or tissue irradiated. To account for this,
tissue weighting factors wT are introduced. They should represent the relative
contribution of an organ or tissue T to the total detriment due to the stochastic
effects resulting from a uniform irradiation of the whole body. The total tissue
weighted equivalent dose is called effective dose and is defined as:
E=
TH T
(3.3)
where HT is the equivalent dose in organ or tissue T.

The sum is performed over all organs and tissues of the human body
considered to be sensitive to the induction of stochastic effects. Recommended
tissue weighting factors are found in ICRP Publication 103 [3.1]. Despite
depending on the sex and age of the person, for the purposes of radiation
protection, the values for tissue weighting factors are taken as constants and are
applicable to the average population.
The use of effective dose has many advantages in practical radiation
protection. Very different exposure situations (e.g. internal and external exposure
by different types of radiation) can be combined and result in a single value, the
effective dose.
3.2.3.3. Committed dose
When radionuclides are taken into the body, the resulting dose is received
throughout the period of time during which they remain in the body. The total
dose delivered during this period of time is referred to as the committed dose and
is calculated as a specified time integral of the rate of receipt of the dose. The
committed equivalent dose is defined as:
H T ( ) =
t0 +
t0
H T (t ) dt (3.4)
where
t0
is the time of intake;
and is the integration time.
79
CHAPTER 3
For workers and adult members of the general public, is taken to be

50 years while for children 70 years is regarded as appropriate.
The committed effective dose is given by:
E( ) =
T H T ( )
(3.5)
3.2.3.4. Operational quantities

For all types of external radiation, the operational quantities for area
monitoring are defined on the basis of a dose equivalent value at a point in the
ICRU sphere. It is a sphere of tissue-equivalent material (30cm in diameter with
a density of 1 g/cm3 and a mass composition of: 76.2% oxygen, 11.1% carbon,
10.1% hydrogen and 2.6% nitrogen). For radiation monitoring, it adequately
approximates the human body in regards to the scattering and attenuation of the
radiation fields under consideration.
The operational quantities for area monitoring defined in the ICRU
sphere should retain their character of a point quantity. This is achieved by
introducing the terms expanded and aligned radiation field in the definition
of these quantities. An expanded radiation field is a hypothetical field in which
the spectral and the angular fluence have the same values at all points of a
sufficiently large volume equal to the values in the actual field at the point of
interest. The expansion of the radiation field ensures that the whole ICRU sphere
is thought to be exposed to a homogeneous radiation field with the same fluence,
energy distribution and direction distribution as at the point of interest of the real
radiation field. If all radiation is aligned in the expanded radiation field so that it
is opposed to a radius vector specified for the ICRU sphere, the aligned and
expanded radiation field is obtained. In this hypothetical field, the ICRU sphere
is homogeneously irradiated from one direction, and the fluence of the field is
the integral of the angular differential fluence at the point of interest in the real
radiation field over all directions. In the expanded and aligned radiation field, the
value of the dose equivalent at any point in the ICRU sphere is independent of
the direction distribution of the radiation in the real radiation field.
For area monitoring, the operational quantity for penetrating radiation is
the ambient dose equivalent, H*(10):
The ambient dose equivalent at a point in a radiation field is the dose
equivalent that would be produced by the corresponding expanded and
aligned field in the ICRU sphere at a depth of 10mm on the radius vector
opposing the direction of the aligned field.
80
For area monitoring, the operational quantity for low-penetrating radiation

is the directional dose equivalent H(0.07, ):
The directional dose equivalent at a point in a radiation field is the dose
equivalent that would be produced by the corresponding expanded field in
the ICRU sphere at a depth of 0.07mm in a specified direction .
The personal dose equivalent Hp(d)is defined as:
The equivalent dose at a depth d in soft tissue below a specified point on
the body.
The relevant depth is d=10mm for penetrating radiations (photon energies
above 15keV), while depths d=0.07mm and d=3mm are used for weakly
penetrating radiations (photon energies below 15keV) in skin and the lens of the
eye, respectively.
3.3. IMPLEMENTATION OF RADIATION PROTECTION
IN A NUCLEAR MEDICINE FACILITY
3.3.1. General aspects
Implementation of radiation protection in a nuclear medicine facility must
fit in with, and be complementary to, the systems for implementing medical
practice in the facility. Radiation protection must not be seen as something
imposed from outside and separate to the real business of providing medical
services and patient care. Most countries have their own radiation protection
legislation and regulatory framework, typically requiring any facility or person
wishing to provide or perform nuclear medicine procedures to have an appropriate
authorization from the radiation protection regulatory body. The requirements to
be fulfilled in order to be granted such an authorization will vary from country to
country, but in general, compliance with the requirements of the BSS would be
expected.
To achieve a high standard of radiation protection, the most important thing
is to establish a safety based attitude in every individual, such that protection and
accident prevention are regarded as a natural part of daily duties. This objective is
basically achieved by education and training, and encouraging a questioning and
learning attitude, but also by a positive and cooperative attitude from the national
authorities and the employer in supporting radiation protection with sufficient
resources, both in terms of personnel and money. A feeling of responsibility
81
CHAPTER 3
can only be achieved if the people involved regard the rules and regulations as
necessary, and are a support to and not a hindrance in their daily work. Every
individual should also know their responsibilities through formal assignment of
duties.
3.3.2. Responsibilities
3.3.2.1. Licensee and employer
The licensee of a nuclear medicine facility, through the authorization issued
by the regulatory body, has the prime responsibility for applying the relevant
national regulations and meeting the conditions of the licence. The licensee may
appoint other people to carry out actions and tasks related to these responsibilities,
but the licensee retains overall responsibility. In particular, the nuclear medicine
physician, the medical physicist, the nuclear medicine technologist, the
radiopharmacist and the radiation protection officer (RPO) all have key roles
and responsibilities in implementing radiation protection in a nuclear medicine
facility, and these are discussed in more detail below.
The BSS need to be consulted for details on all of the requirements for
radiation protection that are assigned to licensees. Employers are also assigned
many responsibilities, in cooperation with the licensee, for occupational radiation
protection. Key responsibilities for the licensee include ensuring that the
necessary personnel (nuclear medicine physicians, medical physicists, nuclear
medicine technologists, radiopharmacists and an RPO) are appointed, and that
the individuals have the necessary education, training and competence to perform
their respective duties. Clear responsibilities for personnel must be assigned; a
radiation protection programme (RPP) must be established and the necessary
resources provided; a comprehensive quality assurance (QA) programme must
be established; and education and training of personnel supported.
3.3.2.2. Nuclear medicine specialist
The general medical and health care of the patient is, of course, the
responsibility of the individual physician treating the patient. However, when the
patient presents in the nuclear medicine facility, the nuclear medicine specialist
has the particular responsibility for the overall radiation protection of the patient.
This means responsibility for the justification of a given nuclear medicine
procedure for the patient, in conjunction with the referring medical practitioner,
and responsibility for ensuring the optimization of protection in the performance
of the examination or treatment.
82
3.3.2.3. Nuclear medicine technologist

The technologist has a key position, and their skill and care to a large extent
determine the optimization of the patients exposure.
3.3.2.4. Radiation protection officer
It is highly recommended that the licensee appoint a person to oversee and
implement radiation protection matters in the hospital. This person is called an
RPO or radiation safety officer. The RPO should have a good theoretical and
practical knowledge of the properties and hazards of ionizing radiation, as well
as protection. In addition, the RPO should possess necessary knowledge of all
the appropriate legislation and codes of practice relating to the uses of ionizing
radiation in the relevant medical area, e.g. nuclear medicine. The RPO, unless
also a qualified medical physicist in nuclear medicine, has no responsibilities for
radiation protection in medical exposure.
3.3.2.5. Medical physicist
The medical physicist is a person who by education and training is
competent to practise independently in one or more of the subfields in
medical physics. For instance, a medical physicist in nuclear medicine should
have a comprehensive knowledge of the imaging equipment used, including
performance specifications, physical limitations of the equipment, calibration,
quality control and image quality. The medical physicist should also be qualified
in handling radiation protection matters associated with nuclear medicine, and
has particular responsibilities for radiation protection in medical exposure,
including the requirements pertaining to imaging (for diagnostic procedures),
calibration, dosimetry and QA. Whenever possible, a medical physicist should
serve as the RPO (seeabove). Other important tasks for the medical physicist
are to be responsible for QA and for the local continuing education in radiation
protection of the nuclear medicine staff and other health professionals.
3.3.2.6. Other personnel
Other personnel that may have responsibilities in radiation protection in
nuclear medicine include radiopharmacists and other staff that may have been
trained to perform special tasks, such as contamination tests or some quality
control tests.
83
CHAPTER 3
3.3.3. Radiation protection programme

The BSS require a licensee (and employer where appropriate) to develop,
implement and document a protection and safety programme commensurate
with the nature and extent of the risks of the practice to ensure compliance with
radiation protection standards. Such a programme is often called an RPP and
each nuclear medicine facility should have one. The RPP for a nuclear medicine
facility is quite complex as it needs to cover all relevant aspects of protection of
the worker, the patient and the general public. The details of such an RPP can be
found in Ref.[3.3].
For an RPP to be effective, the licensee needs to provide for its
implementation, including the resources necessary to comply with the programme
and arrangements to facilitate cooperation between all relevant parties.
3.3.4. Radiation protection committee
An effective way to supervise compliance with the RPP is the formation of
a committee for radiation protection. Since a representative of the management
is usually a member of the radiation protection committee, communication
with the representative may be the most appropriate. The members of the
radiation protection committee should include an administrator representing the
management, the chief nuclear medicine physician, a medical physicist, the RPO,
a nuclear medicine technologist, possibly a nurse for patients undergoing therapy
with radiopharmaceuticals, and a maintenance engineer.
3.3.5. Education and training
According to the BSS, provision must be made to ensure that all personnel
on whom protection and safety depend are appropriately trained and qualified
so that they understand their responsibilities and perform their duties with
appropriate judgement and according to defined procedures. Such personnel
clearly include the nuclear medicine physician (or other medical specialist
wishing to perform nuclear medicine procedures), nuclear medicine technologist,
medical physicist, radiopharmacist and the RPO. However, there are additional
staff that may also need appropriate training, such as nurses working with
radioactive patients and maintenance staff. Details about appropriate levels of
training are given in Ref.[3.3].
84
3.4. FACILITY DESIGN

It is an important task for the medical physicist to be actively involved in
the planning and design of the nuclear medicine facility. Factors that are to be
considered are:
Safety of sources;
Optimization of protection for staff and the general public;
Preventing uncontrolled spread of contamination;
Maintaining low background where most needed;
Fulfilment of national requirements regarding pharmaceutical work.
3.4.1. Location and general layout
The location of the nuclear medicine facility within the hospital or clinic
is not critical, but a few factors need to be considered. It should be readily
accessible, especially for outpatients, who constitute the majority of the patients.
The facility should also be located away from radiotherapy sources and other
strong sources of ionizing radiation such as a cyclotron, which can interfere with
the measuring equipment. Isolation wards for patients treated with radionuclides
should be located outside of the nuclear medicine facility.
The general layout of the nuclear medicine facility should take into account
a possible separation of the work areas and the patient areas. It is also essential to
reduce uncontrolled spread of contamination. This will be achieved by locating
rooms for preparation of radiopharmaceuticals as far away as possible from
rooms for measurements and patient waiting areas. Another important factor is to
reduce the transport of unsealed sources within the facility. The general layout is
from a low activity area close to the entrance to high activity areas at the opposite
end. More details regarding floor planning and additional topics can be found in
the IAEAs Nuclear Medicine Resources Manual [3.4]. It should be borne inmind
that the design of facilities is an important tool in the optimization of protection
of workers and the general public. This is further discussed in Section 3.6.2.
3.4.2. General building requirements
The design of the facility should take into consideration the type of work
to be performed and the radionuclides (and their activity) intended to be used.
The ICRPs concept of categorization of hazard can be used in order to determine
the special needs concerning ventilation and plumbing, and the materials used
in walls, floors and work-benches. The different rooms in the facility will be
categorized as low, medium or high hazard areas.
85
CHAPTER 3
Of special concern in a nuclear medicine facility is the risk of

contamination, and if contamination occurs, the ability to contain it and clean
it up. Therefore, the floors and work-benches should generally be finished in an
impermeable material which is washable and resistant to chemical change, with
all joints sealed. The floor cover should be curved to the wall. The walls should
also be easily cleaned. Chairs and beds used in high hazard areas should be easily
decontaminated. However, some attention has to be given to the comfort of the
patients, for instance in the waiting areas.
Rooms in which unsealed sources, especially radioactive aerosols or gases,
may be produced or handled should have an appropriate ventilation system
that includes a fume hood, laminar air flow cabinet or glove box. It should be
noted that this might also be necessary in the examination room depending on
the radiopharmaceutical used in ventilation scintigraphy. Details regarding fume
hoods, etc. are given in Chapter9.
If the regulatory body allows the release of aqueous waste to the sewer,
a dedicated sink needs to be used, and this needs to be easily decontaminated.
Local rules for the discharge shall be available.
A separate bathroom for the exclusive use by injected patients is
recommended. A sign requesting patients to always sit down, flush the toilet
well and wash their hands should be displayed to lower the risk of contamination
of the floor and to ensure adequate dilution of excreted radioactive materials.
The bathroom should include a sink as a normal hygiene measure and should
be finished in materials that are easily decontaminated. Local rules should be
available for cleaning the toilet. The patient toilet facilities should not be used
by hospital staff as it is likely that the floor, toilet seat and taps will frequently be
contaminated.
Drain-pipes from the nuclear medicine facility should go as directly as
possible to the main building sewer. It should be noted that some countries require
that drain-pipes from a nuclear medicine facility and especially from isolation
wards for patients undergoing radionuclide therapy end up in a delay tank.
3.4.3. Source security and storage
The licensee needs to establish a security system to prevent theft, loss,
unauthorized use or damage to sources. It should be included in all steps from
ordering and delivery of the sources to disposal of spent sources. Only authorized
personnel are permitted to order radionuclides. Routines for delivery and
unpacking shipments should be available, as well as routines for safe handling
and storage of sources. Records of all sources should be kept. The user is always
responsible for the security of sources and, in principle, it should be possible
to identify where an individual source is located or how it has been used, even
86
if it has left the facility in a patient. The regulatory body should promptly be
informed in cases of lost or stolen sources.
When a radioactive source is not in use, it should always be stored. In a
nuclear medicine facility, the sources are generally stored in the room where
preparation of radiopharmaceuticals is undertaken. Storage of sources is further
discussed in Chapter9.
It is necessary to consider the possible consequences of an accidental fire
and to take steps tominimize the risk of this. Careful selection of non-flammable
construction materials when building the storage facility will greatly reduce this
hazard. The storage facility should not be used to hold any highly flammable or
highly reactive materials. Liaison with the local firefighting authority is necessary
and their advice should be sought regarding provision of firefighting equipment
in the vicinity of the radioactive waste store.
3.4.4. Structural shielding
Structural shielding should be considered in a busy facility where large
activities are handled and where many patients are waiting and examined. In a
PET/CT facility, structural shielding is always necessary and the final design will
generally be determined by the PET application because of the high activities
used and because of the high energy of the annihilation radiation. Careful
calculations should be performed to ensure the need and construction of the
barrier. Such calculations should include not only walls but also the floor and
ceiling, and must be made by a qualified medical physicist. Radiation surveys
should always be performed to ensure the correctness of the calculations.
The correct design of protective barriers is of the utmost importance not
only from a protection but also from an economic point of view. If the basic
calculations are wrong, it will become very expensive to correct the mistakes
later when the whole construction is completed. It is, therefore, very important
that a qualified expert, such as a medical physicist, be consulted in the planning
stage.
3.4.5. Classification of workplaces
With regard to occupational exposure, the BSS require the classification of
workplaces as controlled areas or as supervised areas.
In a controlled area, individuals follow specific protective measures to
control radiation exposures. It will be necessary to designate an area as controlled
if it is difficult to predict doses to individual workers or if individual doses
may be subject to wide variations. The controlled area must be delineated and
87
CHAPTER 3
it is convenient to use existing structural boundaries, which should already be

considered at the planning stage of a facility.
A supervised area is any area for which occupational exposure conditions
are predictable and stable. They are kept under review even though specific
additional protective measures and safety provisions are not normally needed.
In a nuclear medicine facility, the rooms for preparation, storage (including
radioactive waste) and injection of the radiopharmaceuticals will be controlled
areas. Owing to the potential risk of contamination, the imaging rooms and
waiting areas for injected patients might also be classified as controlled areas.
The area housing a patient to whom therapeutic amounts of activity have been
given will also be a controlled area. In the case of pure emitters, such as 90Y,
89
Sr or 32P, which are not excreted from the body, the area may not need to be
classified as a controlled area.
3.4.6. Workplace monitoring
Workplace monitoring means checking the facility for the presence of
radiation or radioactive contamination. The two basic types of workplace
monitoring are exposure monitoring and contamination monitoring. Exposure
monitoring (sometimes called area monitoring or radiation surveying) consists
of measuring radiation levels (in microsieverts per hour) at various points using
an exposure meter or survey meter. Contamination monitoring is the search for
extraneous radioactive material deposited on surfaces.
Routine workplace monitoring should be performed at predefined places
in the facility as defined by the RPO. It is an advantage if one member of staff is
appointed to take the measurements. The staff member should be well trained in
handling the instrument. The results should be recorded and investigated if they
exceed the investigation levels predefined by the RPO.
More details regarding workplace monitoring are given in Chapters 9
and 20.
3.4.7. Radioactive waste
The radioactive waste in a nuclear medicine facility comprises many
different types of waste. It may be of high activity, such as a technetium generator,
or of low activity, such as from biomedical procedures or research. In addition,
it may have a long or short half-life and it may be in a solid, liquid or gaseous
form. Radioactive waste needs to be safely managed because it is potentially
hazardous to human health and the environment. Through good practices in the
use of radionuclides, the amount of waste can be significantly reduced but not
eliminated. It is important that safe waste management, in full compliance with
88
all relevant regulations, is considered and planned for at the early stages of any
projects involving radioactive materials. It is the responsibility of the licensee to
provide safe management of the radioactive waste. It should be supervised by the
RPO and local rules should be available.
Containers to allow segregation of different types of radioactive waste
should be available in areas where the waste is generated. The containers must
be suitable for the purpose (volume, shielding, being leakproof, etc.). Each type
of waste should be kept in separate containers that are properly labelled to supply
information about the radionuclide, physical form, activity and external dose rate.
A room for interim storage of radioactive waste should be available. The
room should be locked, properly marked and, if necessary, ventilated. Flammable
waste should be placed separately. It is essential that all waste be properly packed
in order to avoid leakage during storage. Biological waste should be refrigerated
or put in a freezer. Records should be kept, so that the origin of the waste can be
identified.
The final disposal of the radioactive waste produced in the nuclear medicine
facility includes several options: storage for decay and disposal as cleared waste
into the sewage system (aqueous waste), through incineration or transfer to a
landfill site (solid waste), or transfer of sources to the vendor or to a special waste
disposal facility outside of the hospital.
For many of the wastes generated in hospitals, storage for decay is a useful
option because the radionuclides generally have short half-lives. This can be
done in the hospital and may include some treatment of the wastes to ensure
safe storage. Other types of waste containing radionuclides with longer half-lives
must be transferred to a special waste treatment, storage and disposal facility
outside of the hospital. One option is to return the source to the vendor. This is an
attractive option for radionuclide generators and might also be useful for sealed
sources used in a quality control programme. The option of returning the source
should be provided for in the purchase process.
For diagnostic patients there is generally no need for collection of excreta.
Ordinary toilets can be used. For therapy patients, there are different policies in
different countries, either to use separate toilets equipped with delay tanks or an
active treatment system, or to allow the excreta to be released directly into the
sewage system. This is further discussed in Chapter20.
3.5. OCCUPATIONAL EXPOSURE
Detailed requirements for protection against occupational exposure are
given in Section 3 of the BSS, and recommendations on how to meet these
requirements are given in IAEA Safety Guides [3.53.7]. All of these safety
89
CHAPTER 3
standards are applicable to nuclear medicine practice, and in addition Ref.[3.3]

provides further specific advice. A summary of the most relevant issues for
nuclear medicine is given in this section.
3.5.1. Sources of exposure
Exposure of workers may arise from unsealed sources either through
external irradiation of the body or through entry of radioactive substances into the
body. The main precautions required in dealing with external irradiation depend
on the physical characteristics of the emitted radiation and the activity as reflected
by the specific dose rate constant as well as the half-life of the radionuclide.
When a radionuclide enters the body, the internal exposure will depend on factors
such as the physical and chemical properties of the radionuclide, the activity and
the biokinetics.
Every type of work performed in a nuclear medicine facility will
make a contribution to the external exposure of the worker: unpacking
radioactive material, activity measurements, storage of sources, preparation of
radiopharmaceuticals, administration of radiopharmaceuticals, patient handling
and examination, care of the radioactive patient and handling of radioactive
waste. Generally, the yearly effective dose to staff working full time in nuclear
medicine with optimized protection should be well below 5 mSv.
Among the different tasks involved, the highest effective dose is received
from the patient at injection and imaging. The dose rate close to the patient can
be quite high, for instance, 300Sv/h at 0.5 m from a patient who has received
350MBq of 18F.
High equivalent dose to the fingers can be received in preparation and
administration of radiopharmaceuticals, even if proper shielding is used. Injecting
eight patients per day with 400MBq of 99mTc per patient has been reported to
give a mean and maximum equivalent dose to the fingers of 80 and 330 mSv/a,
respectively, even if syringe shields are used. Without shielding, the maximum
equivalent dose will be about 2500 mSv/a.
Higher risk of internal exposure due to contamination is associated with
radioactive spills, animal experiments, emergency surgery of a therapy patient
and autopsy of a therapy patient. However, traces of the radionuclides used in
a nuclear facility can be found almost everywhere, especially on door handles,
taps, some specific equipment and in the patients toilet. Some procedures, such
as ventilation scans, might also cause contamination of both personnel and
equipment. Whole body measurements of workers have revealed an equilibrium
internal contamination of up to 10 kBq of 99mTc, which will result in an effective
dose of ~0.05 mSv/a. Although this is a small fraction of the external exposure,
every precaution must be taken to avoid contamination of the facility.
90
Of special concern is contamination of the skin, since this can result in

extremely high local equivalent doses. For instance, 1 kBq of 18F will result
in an initial equivalent dose rate to the skin of 0.8 mSv/h. The activity on the
hands after elution, preparation and administration of 99mTc radiopharmaceuticals
has been reported to be 0.02200 kBq, which results in an initial skin dose of
0.00550 mSv/h.
3.5.2. Justification, optimization and dose limitation
Nuclear medicine workers have no personal benefit from exposure.
Therefore, justification of occupational exposure must be included in justification
of the nuclear medicine practice itself. The risks in radiation work should not be
greater than for any other similar work. The upper limit of a tolerable risk for the
individual is determined by the dose limits (seeTable3.1). However, through
optimized protection, the incurred effective dose should be further reduced.
Besides facility and equipment design, shielding of sources, handling of sources
as well as personal protective equipment are important in the optimization
of occupational radiation protection. Optimization is also achieved through
education and training, resulting in awareness and involvement in radiation
protection.
From the examples above, it should be clear that the dose limits for
workers can be exceeded if the necessary protective precautions are not taken.
Radiation protection measures must be applied in each step of the work with
radiopharmaceuticals in the nuclear medicine facility, including work with the
patient.
The principal parties responsible for occupational exposure are licensees
and employers, and they should ensure that the exposure is limited and that
protection is optimized. The worker also has responsibilities and must follow
the rules and procedures as well as using the devices for monitoring and the
protective equipment and tools provided, and in all aspects cooperate with the
employer in order to improve the protection standard in the workplace.
3.5.3. Conditions for pregnant workers and young persons
It is generally accepted that the unborn child should be afforded the same
protection level as a member of the general public, meaning that a dose limit
of 1 mSv should be applied once pregnancy is declared. Good operational
procedures should ensure that the radiation doses received by staff working in
nuclear medicine facilities are well below any occupational dose limits. Therefore,
there is generally no need for a pregnant member of staff to change her duties
based on the expected dose to the embryo or fetus. However, removal of pregnant
91
CHAPTER 3
women from work in laboratories where large quantities of radionuclides are

prepared and administered, and from nursing teams responsible for patients who
have been treated with radionuclides should be considered. These staff members
could receive a dose to the embryo or fetus comparable with the public dose limit
over the period of the pregnancy. Since all doses should be reduced whenever
possible, some supervisors will consider it prudent to reassign pregnant staff to
non-radiation duties if this is possible. Many nuclear medicine facility managers
would also accept requests from women to be reassigned to other duties for
reasons beyond radiation protection. Previous personal monitoring results can
help guide any decisions, noting that the dose to the fetus from external radiation
is not likely to exceed 25% of the personal dosimeter measurement.
According to the BSS, no person under the age of 16 years is to be
subjected to occupational exposure, and no person under the age of 18 years is to
be allowed to work in a controlled area unless supervised and then only for the
purpose of training.
3.5.4. Protective clothing
Suitable personal protective clothing should be provided for the use of all
persons employed in work in controlled areas. The protective clothes should be
adequate to prevent any contamination of the body of the worker for whom it is
provided and should include gloves, laboratory coats, safety glasses and shoes or
overshoes, as well as caps and masks for aseptic work.
A question frequently asked is whether lead aprons are useful for nuclear
medicine work. Wearing a lead apron at all times will reduce the effective dose
by a factor of about two. It is, therefore, a matter of judgement whether this dose
reduction compensates for the effort of wearing an apron. In some hospitals, lead
aprons are used in the case of prolonged injections and high activity.
3.5.5. Safe working procedures
The safety of the work in nuclear medicine is based on facility design as
well as on the use of protective clothing and the use of protective equipment
and tools as discussed above. These measures together with working procedures
aimed tominimize external exposure, risk of contamination and spread of
contamination, will optimize protection of workers. Work with unsealed sources
should always be supported by written local rules.
In order tominimize the risk of contamination in handling
radiopharmaceuticals, clean operation conditions and good laboratory practice
should be adopted, and protective clothing used. The work area should be
kept tidy and free from articles not required for work. It should be monitored
92
periodically and be cleaned often enough to ensureminimal contamination.

No food or drink, cosmetic or smoking materials, crockery or cutlery should
be brought into an area where unsealed radioactive substances are used. They
should not be stored in a refrigerator used for unsealed radioactive substances.
Handkerchiefs should never be used in these areas.
All manipulation for preparation, dispensing and administration of
radioactive materials should be carried out in such a way that the spread
of contamination isminimized. That includes preparing and dispensing
radiopharmaceuticals over a drip tray covered with absorbing paper as well as
using absorbing compresses at administrations. Any spills of radioactive material
should be immediately covered with absorbent material to prevent the spread of
material. If the spill cannot be cleaned up immediately, it must be marked to warn
other personnel of its location. Decontamination of the area must begin as soon
as possible.
When wearing gloves which may be contaminated, unnecessary contact
with all other objects should be avoided. Gloves should be removed and disposed
of in the radioactive waste bin as soon as work with radioactive substances is
finished.
After finishing work with the potential for contamination, the protective
clothing should be removed and placed in an appropriate container. Hands should
be washed and monitored.
In order tominimize external exposure, the three fundamental measures
of protection should be applied: time, distance and shielding. As far as possible,
the time of exposure should be as short as possible. Of course, this is important
in work where high exposure rates can be expected, such as in the preparation of
radiopharmaceuticals. However, limiting exposure time should not compromise
the quality of work or the use of other protective measures.
Direct handling of vials, syringes or other sources which produce a
significant radiation field is not recommended. Forceps or tongs should be used
to reduce the radiation exposure by increasing the distance between the source
and the hands. Properly designed vial and syringe shields must be used wherever
practicable. In cases where unshielded sources are handled or the exposure time
is prolonged, the work should be performed behind a properly designed lead
glass shield or similar type of protective barrier.
Radioactive waste should not be stored in the work area but transferred to a
separate radioactive waste storage room as soon as possible.
A patient undergoing a nuclear medicine imaging study is a source
of radiation exposure and contamination. Contact with these patients by
nursing staff presents little hazard, as the radiation dose rate is quite low, and
accumulated dose to any single individual would not be significant. However, for
nuclear medicine staff that spend a great deal of time in the immediate vicinity of
93
CHAPTER 3
these patients, the accumulated radiation dose can be significant. These workers
should, whenever possible, maximize their distance from the patient and spend as
little time as possible in close proximity to the patient.
In summary, the following protective approaches can reduce external
exposure significantly:
For preparation and dispensing of radiopharmaceuticals, working behind a
lead glass bench shield, and using shielded vials and syringes;
For administration of radiopharmaceuticals to patients, using lead aprons in
the case of prolonged injection and high activity, and using a syringe shield;
During examinations, when the distance to the patient is short, using a
movable transparent shield.
3.5.6. Personal monitoring
The licensee and employer have the joint responsibility to ensure that
appropriate personal monitoring is provided to staff. This normally means that
the RPO would specify which workers need to be monitored routinely, the type
of monitoring device to be used and the body position where the monitor should
be worn, bearing inmind that some countries may have specific regulatory
requirements on these issues. Further, the regulatory body is likely to have
specified the monitoring period and the time frame for reporting monitoring
results.
Staff to be monitored in a nuclear medicine facility should include all those
who work routinely with radionuclides or with the patients who have received
administrations of radiopharmaceuticals. This will include nursing staff who
either work routinely in nuclear medicine or nurse patients who have received
radionuclide therapy and staff dealing with excreta from radionuclide therapy.
Monitoring would not normally be extended to those that come into occasional
contact with nuclear medicine patients.
There are several types of external personal dosimetry systems and the
system to use is dependent on national or local conditions. In many countries,
the service is centralized to the regulatory body or provided through third party
personal dosimetry providers. Occasionally, some large hospitals have their own
personal dosimetry service. In all cases, the dosimetry provider must be approved
by the regulatory body.
Finger monitoring should be carried out occasionally on staff that regularly
prepare and administer radioactive substances to patients, and also when
setting up an operation which requires the routine handling of large quantities
of radionuclides. After handling unsealed radionuclides, the hands should be
monitored. It may, therefore, be convenient to mount a suitable contamination
94
monitor near the sink where hands are washed. Care should be taken to ensure
that the monitor itself does not become contaminated. In high background areas,
it will be necessary to shield the detector, and it may be convenient to have a foot
or elbow operated switch to activate the monitor.
Monitoring for internal contamination is rarely necessary in nuclear
medicine on radiation protection grounds but it may be useful in providing
reassurance to staff. The circumstances in which internal monitoring becomes
advisable are those where staff use significant quantities of 131I for thyroid therapy.
They should be included in a programme of thyroid uptake measurements.
In other circumstances where it is necessary to assess the intake of emitting
radionuclides (e.g. after a serious incident), the use of a whole body counter may
be appropriate. Such equipment should be available at national referral centres.
The possible use of an uncollimated gamma camera should also be considered.
Sometimes, a more detailed monitoring survey may be indicated if staff
doses have increased (or it is anticipated that they may do so in the future) as a
result of either the introduction of new examinations or procedures, or a change
in the nuclear medicine facilitys equipment. The RPO should decide who should
be monitored and at which monitoring sites.
Individual monitoring results must be analysed and records must be kept. It
is vital that the individual monitoring results are regularly assessed and the cause
of unusually high dosimeter readings should be investigated by the RPO, with
ensuing corrective actions where appropriate. The administrative arrangements,
the scope and nature of the individual monitoring records, and the length of time
for which records have to be kept may differ among countries.
3.5.7. Monitoring of the workplace
The BSS require licensees to develop programmes for monitoring the
workplace. Such programmes are described in Section 3.4.6 and in Chapters 9
and 20.
3.5.8. Health surveillance
According to the BSS, the licensee needs to make arrangements for
appropriate health surveillance in accordance with the rules established by
the national regulatory body. The primary purpose of health surveillance is to
assess the initial and continuing fitness of employees for their intended tasks.
The health surveillance programme should be based on the general principles of
occupational health.
No specific health surveillance related to exposure to ionizing radiation is
necessary for staff involved in nuclear medicine procedures. Only in the case of
95
CHAPTER 3
overexposed workers at doses much higher than the dose limits would special
investigations involving biological dosimetry and further extended diagnosis and
medical treatment be necessary.
Counselling should be available to workers such as women who are
or may be pregnant, individual workers who have or may have been exposed
substantially in excess of dose limits and workers who may be worried about
their radiation exposure.
3.5.9. Local rules and supervision
According to the BSS, employers and licensees must, in consultation with
the workers or through their representatives:
Establish written local rules and procedures necessary to ensure adequate
levels of protection and safety for workers and other persons;
Include in the local rules and procedures the values of any relevant
investigation level or authorized level, and the procedure to be followed in
the event that any such value is exceeded;
Make the local rules and procedures, the protective measures and safety
provisions known to those workers to whom they apply and to other persons
who may be affected by them;
Ensure that any work involving occupational exposure be adequately
supervised and take all reasonable steps to ensure that the rules, procedures,
protective measures and safety provisions be observed.
These local rules should include all working procedures involving unsealed
sources in the facility such as:
Ordering radionuclides;
Unpacking and checking the shipment;
Storage of radionuclides;
General rules for work in controlled and supervised areas;
Preparation of radiopharmaceuticals;
Personal and workplace monitoring;
In-house transport of radionuclides;
Management of radioactive waste;
Administration of radiopharmaceuticals to the patients;
Protection issues in patient examinations and treatments;
Routine cleaning of facilities;
Decontamination procedures;
Care of radioactive patients.
96
It is the responsibilty of the licensee of the nuclear medicine facility to

ensure that local rules are established, maintained and continually reviewed. The
RPO would have significant involvement in this process.
3.6. PUBLIC EXPOSURE
3.6.1. Justification, optimization and dose limitation
According to the BSS, public exposure is exposure incurred by members of
the public from radiation sources, excluding any occupational or medical exposure.
The three ICRP principles described in Section 3.2.1 apply to public
exposure arising from the practice of nuclear medicine. Just as for occupational
exposure, the justification of public exposure is based on the justification of the
practice of nuclear medicine. The exposure of the general public is ultimately
restricted by the application of dose limits (seeTable3.1), but in the first instance
the application of the principle of optimization of protection ensures that public
doses will be ALARA.
The licensee is responsible for controlling public exposure arising from a
nuclear medicine facility. The presence of members of the public in or near the
nuclear medicine facility needs to be considered when designing the shielding
and flow of persons in the facility.
The sources of exposure of the general public are primarily the same as for
workers. Hence, the use of structural shielding and the control of sources, waste
and contamination are fundamental to controlling exposure of the public. There
are, however, some additional situations that need special consideration. These
include the release of patients examined or treated with radiopharmaceuticals.
3.6.2. Design considerations
The general layout of the nuclear medicine facility should take into account
the protection of members of the public. The areas for storage and preparation of
radiopharmaceuticals must be well separated from public areas such as waiting
rooms. The movement of radionuclides must beminimized. For example, the
room for preparation and dispensing of radiopharmaceuticals and the room for
administration should be adjacent and connected by a pass through. Areas where
significant activities of radionuclides are present must be appropriately shielded.
Access must be restricted so that members of the public are not allowed into
controlled areas. Radioactive waste must be stored in a secure location away from
areas accessible to the public. Since a patient still waiting for administration of
the radiopharmaceutical is regarded as a member of the public, separate waiting
97
CHAPTER 3
rooms and toilets for injected and not injected patients should be considered in
order tominimize both external exposure and the spread of contamination.
3.6.3. Exposure from patients
Every precaution must be taken to ensure that the doses received by
individuals who come close to a patient or who spend some time in neighbouring
rooms remain below the dose limit for the public and below any applicable dose
constraint. For almost all diagnostic procedures, the maximum dose that could be
received by another person due to external exposure from the patient is a fraction
of the annual public dose limit and it should not normally be necessary to issue any
special radiation protection advice to the patient. One exception is restrictions on
breast-feeding a baby, which will be further discussed in Section 3.7.2.4. Another
exception is an intensive use of positron emitters which may require structural
shielding based on the exposure of the public as discussed above (Section 3.4.4).
For patients who have undergone radionuclide therapy, specific advice should be
given regarding restrictions on their contact with other people. This is discussed
separately in Chapter20.
3.6.4. Transport of sources
One possible source of exposure of the general public is transport of
sources. It is performed both inside and outside the nuclear medicine facility.
Inside the facility, the transport includes distribution of the radioactive sources
from the storage area to where it will be used. Such transport should be limited
as far as possible by the facility design. The transport that takes place should
be performed according to optimized radiation protection conditions as given by
local rules.
The transport of radioactive sources to and from the nuclear medicine
facility should follow the internationally accepted IAEA Regulations for the Safe
Transport of Radioactive Material [3.8]. These Regulations include basic rules
for the transport itself and regulations about the shape and labelling of packages.
In general, the package is built in several parts. It should be mechanically
safe and reduce the effect of potential fire and water damage. The package should
be labelled with a sign. There are three different labels: IWhite, IIYellow and
IIIYellow. In all cases, the radionuclide and its activity should be specified. The
label gives some indication of the dose rate D at the surface of the package:
Category IWhite
D 0.005 mSv/h
Category IIYellow 0.005 < D 0.5 mSv/h
Category IIIYellow 0.5 < D 2 mSv/h
98
A more exact figure of the radiation around the package is given by the
transport index which is the maximum dose rate (mSv/h) at a distance 1 m from
the surface of the package multiplied by a hundred.
3.7. MEDICAL EXPOSURE
The detailed requirements given in Section 3 of the BSS are applicable to
medical exposure in nuclear medicine facilities. Furthermore, Ref.[3.9] describes
strategies to involve organizations outside the regulatory framework, such as
professional bodies (nuclear medicine physicians, medical physicists, nuclear
medicine technologists, radiopharmacists), whose cooperation is essential to
ensure compliance with the BSS requirements for medical exposures. Examples
that may illustrate this point include the adoption of protocols for calibration
of unsealed sources and for QA and for reporting accidental medical exposure.
Reference [3.3] provides further specific advice. A summary of the most relevant
issues for nuclear medicine is given in this section.
3.7.1. Justification of medical exposure
The BSS state that:
Medical exposures shall be justified by weighing the expected diagnostic
or therapeutic benefitsthat they yield against the radiation detriment
that they might cause, with account taken of the benefits and the risks of
available alternative techniques that do not involve medical exposure.
The principle of justification of medical exposure should not only be applied
to nuclear medicine practice in general but also on a case by case basis, meaning
that any examination should be based upon a correct assessment of the indications
for the examination, the actual clinical situation, the expected diagnostic and
therapeutic yields, and the way in which the results are likely to influence the
diagnosis and the medical care of the patient. The nuclear medicine specialist
has the ultimate responsibility for the control of all aspects of the conduct and
extent of nuclear medicine examinations, including the justification of the given
procedure for a patient. The nuclear medicine specialist should advise and make
decisions on the appropriateness of examinations and determine the techniques
to be used. In justifying a given diagnostic nuclear medicine procedure, relevant
international or national guidelines should be taken into account.
Any nuclear medicine procedure that occurs as part of a biomedical
research project (typically as a tool to quantify changes in a given parameter
99
CHAPTER 3
under investigation) is considered justified if the project has been approved by an

ethics committee.
3.7.2. Optimization of protection
The principle of optimization of protection is applied to nuclear medicine
procedures that have been justified, and can be summarized as follows. For
diagnostic nuclear medicine procedures, the patient exposure should be
theminimum necessary to achieve the clinical purpose of the procedure,
taking into account relevant norms of acceptable image quality established by
appropriate professional bodies and relevant diagnostic reference levels (DRLs).
For therapeutic nuclear medicine procedures, the appropriate
radiopharmaceutical and activity are selected and administered so that the activity
is primarily localized in the organ(s) of interest, while the activity in the rest of
the body is kept ALARA.
The implementation of optimization of protection for patients in
nuclear medicine is quite complex and includes equipment design, choice of
radiopharmaceutical and activity, procedure considerations, DRLs, calibration,
clinical dosimetry and QA, as well as special considerations for children,
pregnant women and lactating women. This is further discussed in the following
sections.
3.7.2.1. Administered activity and radiopharmaceuticals
For diagnostic procedures, it is necessary for the nuclear medicine specialist
in cooperation with the medical physicist to determine the optimum activity to
administer in a certain type of examination, taking the relevant DRL (seebelow)
into account. For any given procedure used on an individual patient, the optimum
activity will depend on the body build and weight of the patient, the patients
metabolic characteristics and clinical condition, the type of equipment used, the
type of study (static, dynamic, tomographic) and the examination time.
For a given type of imaging equipment, the diagnostic value of the
information obtained from an examination will vary with the amount of
administered activity. There is a threshold of administered activity below which
no useful information can be expected. Above this level, the diagnostic quality
will increase steeply with increasing activity. Once an acceptable image quality
has been reached, a further increase of the administered activity will only increase
the absorbed dose and not the value of the diagnostic information.
It should also be noted that limiting the administered activity below the
optimum, even for well intentioned reasons, will usually lead to a poor quality
of the result which may cause serious diagnostic errors. It is very important to
100
avoid failure to obtain the required diagnostic information; failure would result
in unnecessary (and, therefore, unjustified) irradiation and may also necessitate
repetition of the test.
If more than one radiopharmaceutical can be used for a procedure,
consideration should be given to the physical, chemical and biological properties
for each radiopharmaceutical, so as tominimize the absorbed dose and other risks
to the patient while at the same time providing the desired diagnostic information.
Other factors affecting the choice include availability, shelf life, instrumentation
and relative cost. It is also important that the radiopharmaceuticals used are
received from approved manufacturers and distributors, and are produced
according to national and international requirements. This is a requirement also
for in-house production of radiopharmaceuticals for PET studies.
The activity administered to a patient should always be determined and
recorded. Knowing the administered activity makes it possible to estimate the
absorbed dose to different organs as well as the effective dose to the patient.
Substantial reduction in absorbed dose from radiopharmaceuticals can be
achieved by simple measures such as hydration of the patient, use of thyroid
blocking agents and laxatives.
3.7.2.2. Optimization of protection in procedures
The nuclear medicine procedure starts with the request for an examination or
treatment. The request should be written and contain basic information about the
patients condition. This information should help the nuclear medicine specialist
to decide about the most appropriate method to use and to decide how urgent
the examination is. The patient should then be scheduled for the examination or
treatment and be informed about when and where it will take place. Some basic
information about the procedure should also be given, especially if it requires
some preparation of the patient, such as fasting. These initial measures require
an efficient and reliable administrative system. In parallel to these routines, the
nuclear medicine facility has to ensure that the radiopharmaceutical to be used is
available at the time of the scheduled procedure.
When the patient appears in the nuclear medicine facility, they should
be correctly identified using the normal hospital or clinic routines. The patient
should be informed about the procedure and have the opportunity to ask questions
about it. A fully informed and motivated patient is the basis for a successful
examination or treatment. Before the administration of the radiopharmaceutical,
the patient should be interviewed about possible pregnancy, small children at
home, breast-feeding and other relevant questions which might have implications
for the procedure. Before administration, the technologist or doctor should
check the request and ensure that the right examination or treatment is scheduled
101
CHAPTER 3
and that the right radiopharmaceutical and the right activity are dispensed. If
everything is in order, the administration can proceed. The administered activity
should always be recorded for each patient.
While most adults can maintain a required position without restraint
or sedation during nuclear medicine examinations, it may be necessary to
immobilize or sedate children, so that the examination can be completed
successfully. Increasing the administered activity to reduce the examination time
is an alternative that can be used in elderly patients with pain.
Optimization of protection in an examination means that equipment should
be operated within the conditions established in the technical specifications, thus
ensuring that it will operate satisfactorily at all times, in terms of both the tasks to
be accomplished and radiation safety. More details are given in Chapters 8 and 15.
Particular procedural considerations for children, pregnant women and lactating
women are given in the following subsections.
Optimization of protection in radionuclide therapy means that a correctly
calculated and measured activity should be administered to the patient in order
to achieve the prescribed absorbed dose in the organ(s) of interest, while the
radioactivity in the rest of the body is kept as low as reasonably achievable.
Optimization also means using routines to avoid accidental exposures of the
patient, the staff and members of the general public. Radionuclide therapy is
further discussed in Chapter20.
The availability of a written manual of all procedures carried out by the
facility is highly desirable. The manual should regularly be revised as part of a
QA programme.
3.7.2.3. Pregnant women
Special consideration should be given to pregnant women exposed to
ionizing radiation due to the larger probability of inducing radiation effects in
individuals exposed in utero compared to exposed adults. As a basic rule, it is
recommended that diagnostic and therapeutic nuclear medicine procedures
of women likely to be pregnant be avoided unless there are strong clinical
indications.
In order to avoid unintentional irradiation of the unborn child, a female of
childbearing age should be evaluated regarding possible pregnancy or a missed
period. This should be done when interviewing and informing the woman prior to
the examination or treatment. It is also common to place a poster in the waiting
area requesting a woman to notify the staff if she is or thinks she is pregnant.
If the patient is found not to be pregnant without any doubt, the examination
or treatment can be performed as planned. If pregnancy is confirmed,
careful consideration should be given to other methods of diagnosis or to the
102
postponement of the examination until after delivery. If, after consultation

between the referring physician and the nuclear medicine specialist, these options
are not feasible, then the examination should be performed, but the process of
optimization of protection needs to also consider protection of the embryo/fetus.
In order to reduce the fetal dose, it may sometimes be possible to reduce
the administered activity and acquire images for longer times, but great care must
be taken not to compromise the quality of the result. After the administration of
radiopharmaceuticals, frequent voiding should be ensured tominimize exposure
from the bladder. This contribution to the fetal dose can be further reduced by
administering the radiopharmaceutical when the bladder is partially filled, rather
than immediately after voiding.
Of special concern is also the use of CT in PET/CT or SPECT/CT
examinations. Routine diagnostic CT examinations of the pelvic region with and
without contrast injection can lead to a dose of 50 mSv to the uterus which is
assumed to be equivalent to the fetal dose in early pregnancy. It is important to
use low dose CT protocols and to reduce the scanning area to aminimum when
PET/CT or SPECT/CT scanning is indicated in a pregnant patient.
Pregnant women should not be subject to therapy with a radioactive
substance unless the application is life-saving. Following treatment with a
therapeutic activity of a radionuclide, female patients should be advised to avoid
pregnancy for an appropriate period. More details are given in Ref.[3.3].
If the fetal dose is suspected to be high (e.g. >10 mSv), it should be carefully
determined by a qualified medical physicist and the pregnant woman should be
informed about the possible risks. The same procedure should be applied in the
case of an inadvertent exposure, which can be incurred by a woman who later
was found to have been pregnant at the time of the exposure or in emergency
situations.
Exposure of a pregnant patient at a time when the pregnancy was not known
often leads to her apprehension because of concern about the possible effects on
the fetus. It may lead to a discussion regarding termination of pregnancy due to
the radiation risks. Many misunderstandings and lack of knowledge, also among
physicians, have probably resulted in unnecessary termination of pregnancies.
It is generally considered that for a fetal dose of less than 100 mGy, as in most
diagnostic procedures, termination of pregnancy is not justified from the point
of radiation risks. At higher doses, individual circumstances should be taken into
account. This is an ethical issue and the national authorities should give guidance.
3.7.2.4. Lactating women
When nuclear medicine examinations are requested for women who are
breast-feeding, they present a potential radiation hazard to the baby. This is due
103
CHAPTER 3
to uptake of some radiopharmaceuticals in breast tissue followed by excretion

into the breast milk. The dose to the baby depends on various factors such as the
radiopharmaceutical, the amount of milk and the time between the administration
of the radiopharmaceutical to the mother and the feeding of the child. The mother
also represents a source of external exposure and contamination when feeding
or cuddling the baby. The dose will depend on the time the child is held, the
distance from the mothers body and personal hygiene. Some restrictions on
breast-feeding and advice to the mother are necessary in order tominimize the
exposure of the baby to an acceptable level. The baby is a member of the public
and a typical constraint on the dose from a single source of exposure (in this case,
per episode) is 0.3 mSv.
Before a nuclear medicine examination or therapy with radionuclides,
the woman should be asked, orally or in writing, whether she is breast-feeding
a child. A notice requesting the patient to inform the staff about breast-feeding
should also be prominently displayed in the waiting area. If the answer is yes,
consideration should be given as to whether the examination or treatment could
reasonably be delayed until she has ceased breast-feeding. If not, advice about
restriction of breast-feeding dependent on the diagnostic or therapeutic procedure
should be given to the patient.
It is the responsibility of the nuclear medicine specialist in cooperation with
the medical physicist to establish local rules regarding breast-feeding and close
contact between the mother and the child after a nuclear medicine examination or
treatment. The rules should be based on recommendations given by international
and national authorities as well as professional organizations. Some guidance is
found in Ref.[3.3].
3.7.2.5. Children
Optimization of protection for an examination of a child is basically an
optimization of the administered activity. There are several approaches to the
problem of how to calculate the administered activity for children. It should be
theminimum consistent with obtaining a diagnostic result. As this is the same
principle which is applied to adult doses, the normal activity administered to
adults should be used as a guide, bearing inmind that the average adult body
weight is 70 kg. For children or young persons, body weight should always
be measured and the adult administered activity should then be scaled down.
Opinions differ as to how the scaling should be achieved. Simply reducing the
activity in proportion to body weight may, in some types of investigation, result
in inadequate image quality. Another method is based on the principle of scaling
in proportion to body surface area. This approach should give the same image
count density as that for an adult patient, although the effective dose is higher. As
104
a general guide, activities less than 10% of the normal adult activity should not
be administered.
In hybrid imaging, the CT protocol should be optimized by reducing the
tube currenttime product (mAs) and tube potential (kV) without compromising
the diagnostic quality of the images. Careful selection of slice width and pitch as
well as scanning area should also be done. It is important that individual protocols
based on the size of the child are used. The principles behind such protocols
should be worked out by the medical physicist and the responsible specialist.
Since the examination times in nuclear medicine examinations are quite
long, there may be problems in keeping the child still during the examination.
Even small body motions can severely interfere with the quality of the
examination and make it useless. There are several methods of mechanical
support to fasten the child. Drawing the childs attention to something else such
as a television programme can also be useful for older children. Sometimes, even
sedation or general anaesthesia may be necessary.
3.7.2.6. Calibration
The licensee of a nuclear medicine facility needs to ensure that a dose
calibrator or activity meter is available for measuring activity in syringes or vials.
The validity of measurements should be ensured by regular quality control of
the instrument, including periodic reassessment of its calibration, traceable to
secondary standards.
3.7.2.7. Clinical (patient) dosimetry
The licensee of a nuclear medicine facility should ensure that appropriate
clinical dosimetry by a medical physicist is performed and documented. For
diagnostic nuclear medicine, this should include representative typical patient
doses for common procedures. For therapeutic nuclear medicine, this needs to
be for each individual patient, and includes absorbed doses to relevant organs or
tissues.
3.7.2.8. Diagnostic reference levels
Many investigations have shown a large spread of administered activities
for a certain type of diagnostic nuclear medicine examination between different
hospitals within a country, even if the equipment used is similar in performance.
Even though no dose limits are applied to medical exposure, the process of
optimization should result in about the same administered activity for the same
type of examination and for the same size of patient.
105
CHAPTER 3
The concept of a DRL provides a tool for the optimization of protection

in medical exposure. In the case of nuclear medicine, the DRL is given as
administered activity for a certain type of examination and for a normal sized
patient. DRLs are aimed to assist in the optimization of protection by helping to
avoid unnecessarily high activities to the patient or too low activities to provide
useful diagnostic information. DRLs are normally set at the national level as a
result of consultation between the health authority, relevant professional bodies
and the radiation protection regulatory body.
3.7.2.9. Quality assurance for medical exposures
The BSS require the licensee of the nuclear medicine facility to have a
comprehensive programme of QA for medical exposures. The programme needs
to have the active participation of the medical physicists, nuclear medicine
specialists, nuclear medicine technologists and radiopharmacists, and needs to
take into account principles established by international organizations, such as
the WHO and PAHO, and relevant professional bodies.
The programme of QA for medical exposures should be complementary to
and part of the wider programme of QA for radiation protection the latter also
including occupational and public exposure. In turn, this programme needs to be
part of and harmonized with the nuclear medicine facilitys quality management
system. Section 3.9 discusses the wider QA programme, while the remainder of
this subsection deals with some aspects of the programme as it applies to medical
exposures.
The programme of QA for medical exposures should include:
Measurements by, or under the oversight of, a medical physicist of the
physical parameters of medical radiological equipment at the time of
acceptance and commissioning prior to clinical use on patients, periodically
thereafter, and after any major maintenance that could affect patient
protection;
Implementation of corrective actions if measured values of the physical
parameters are outside established tolerance limits;
Verification of the appropriate physical and clinical factors used in patient
diagnosis or treatment;
Records of relevant procedures and results;
Periodic checks of the appropriate calibration and conditions of operation
of dosimetry and monitoring of equipment.
In addition, the licensee needs to ensure that there are regular and
independent audits of the programme of QA for medical exposures, their
106
frequency depending on the complexity of the nuclear medicine procedures

performed and the risks involved.
The above indicates, among other actions, the need for quality control tests
on the equipment. More details regarding quality control of equipment used in
diagnosis will be found in other chapters of this book.
3.7.3. Helping in the care, support or comfort of patients
Certain patients, such as children, the elderly or the infirm, may have
difficulty during a nuclear medicine procedure. Occasionally, people knowingly
and voluntarily (other than in their employment or occupation) may volunteer
to help in the care, support or comfort of patients. In such circumstances, the
dose to these persons (excluding children and infants) should be constrained
so that it is unlikely that it will exceed 5 mSv during the period of a patients
diagnostic examination or treatment. The dose to children visiting patients who
have ingested radioactive materials should be similarly constrained to less than
1 mSv. Special concern should be given to members of the family of a patient
who has received radionuclide therapy. This is further discussed in Chapter20.
Sometimes, a nurse escorting a patient to the nuclear medicine facility is
asked to provide assistance during a procedure. Any resultant exposure should
be regarded as occupational, and the nurse should have received education and
training on this role.
3.7.4. Biomedical research
The exposure of humans for biomedical research is deemed not to be
justified unless it is in accordance with the provisions of the Helsinki Declaration
[3.10] and follows the guidelines for its application prepared by the Council for
International Organizations of Medical Sciences [3.11]. It is also subject to the
approval of an ethics committee.
The use of radioactive trace substances is common in biomedical research.
Diagnostic nuclear medicine procedures may be part of a biomedical research
project, typically as a means for quantifying changes in a given parameter
under investigation or assessing the efficacy of a treatment under investigation.
An exposure as part of biomedical research is treated on the same basis as a
medical exposure and, therefore, is not subject to dose limits. However, in all
investigations involving exposure of humans, a careful estimation of the radiation
dose to the volunteer should be made. The associated risk should then be weighed
against the benefit for the patient or society. Recommendations are given by the
ICRP. The BSS require the use of dose constraints, on a case by case basis, in the
process of optimization.
107
CHAPTER 3
3.7.5. Local rules

The management of patients in the nuclear medicine facility should be
supported by written local rules covering all procedures that may affect medical
exposure. These local rules should be signed by the responsible person and
known to every member of the staff and should include:
Routines for patient identification and information;
Prescribed radiopharmaceutical and activity for adults and children for
different types of examination, including methods used to adjust the activity
to the single patient and routes of administration;
Management of patients that are pregnant or might be pregnant;
Management of breast-feeding patients;
Routines for safe preparation and administration of radiopharmaceuticals
including activity measurements;
Procedures in case of misadministration of the radiopharmaceutical;
Detailed procedure manuals for every type of examination including
handling of equipment.
3.8. POTENTIAL EXPOSURE
3.8.1. Safety assessment and accident prevention
Unintended and accidental exposure may occur due to equipment failure,
human error or a combination of both. Although such events can be identified
by a careful safety assessment, their details and the time of occurrence cannot be
predicted. These exposures are called potential exposures. It is the responsibility
of the licensee to take measures in order to prevent such events as far as possible
and, in case they occur, mitigate their consequences.
According to the BSS, the licensee needs to conduct a safety assessment
applied to all stages of the design and operation of the nuclear medicine facility,
and present the report to the regulatory body if required. The safety assessment
needs to include, as appropriate, a systematic critical review of identification of
possible events leading to unintended or accidental exposure. In practice, this
means that all procedures in which unsealed sources are involved in the work
should be listed and for every procedure it should be asked what can go wrong.
Some examples are given in Table3.2.
108
TABLE3.2. EXAMPLES OF WHAT CAN POTENTIALLY GO WRONG IN A

NUCLEAR MEDICINE FACILITY
Procedure and involvement
What can go wrong?
Patients involved
Request and scheduling
Wrong patient scheduled
Identification at arrival
Wrong patient identified
Information
Missed pregnancy or breast-feeding
Administration of
radiopharmaceutical
Misadministration (wrong patient, wrong

activity, wrong radiopharmaceutical)
Waiting
Contamination of waiting area (vomiting,

incontinence)
Examination
Inconclusive due to contamination,

equipment and/or software failure
Workers involved
Ordering of sources
Unauthorized ordering
Receipt and unpacking of shipments
Damage to package, contamination
Storage of sources
Unshielded sources, high dose rates,

loss of sources
Preparation and administration of

radiopharmaceutical
High doses recorded, contamination of

workers and facilities
Handling of radioactive waste
Contamination of workers and facilities
General public involved

Storage of sources
Loss of sources
Handling of sources
Contamination of facility
Radioactive waste
Loss of sources, contamination of facilities
Radioactive patient
Escape of hospitalized patient, medical

emergency, death of patient
Undertaking a safety assessment requires using ones imagination to try to

define an event that could result in a potential exposure, even if the event has
never occurred before. For instance, what should be done if a patient who just
received 15 GBq of 131I escapes from the isolation ward and the hospital and is
seriously injured in a road accident?
If an unintended or accidental medical exposure occurs, the licensee is
required to determine the patient doses involved, identify any corrective actions
109
CHAPTER 3
needed to prevent recurrence and implement the corrective measures. There may
be a requirement to report the event to the regulatory body.
A well established RPP is fundamental in accident prevention together with
a high level of safety culture in the organization and among the people working
in a nuclear medicine facility. The content of an RPP as well as the importance of
well established working procedures in order to protect patients, workers and the
general public have been discussed in the sections above. It should be stressed
that documentation of the procedures used in the facility is also important in
accident prevention. Other important factors are a well working QA programme
and a programme for continuing education and training which includes not only
the normal practices, but also accidental situations and lessons learned from
accidents.
3.8.2. Emergency plans
According to the BSS, the licensee needs to prepare emergency procedures
on the basis of events identified by the safety assessment. The procedures should
be clear, concise and unambiguous, and need to be posted visibly in places where
their need is anticipated. An emergency plan needs to, as aminimum, list and
describe:
Predictable incidents and accidents, and measures to deal with them;
The persons responsible for taking actions, with full contact details;
The responsibilities of individual personnel in emergency procedures
(nuclear medicine physicians, medical physicists, nuclear medicine
technologists, etc.);
Equipment and tools necessary to carry out the emergency procedures;
Training and periodic drills;
The recording and reporting system;
Immediate measures to avoid unnecessary radiation doses to patients, staff
and the public;
Measures to prevent access of persons to the affected area;
Measures to prevent spread of contamination.
The most likely accident in a nuclear medicine facility is contamination
of workers, patients, equipment and facilities. It can range from small to very
large spillages of radioactivity, for example, serious damage to the technetium
generator or spillage of several gigabecquerels of 131I. The procedures of cleaning
up a small amount of contamination should be known and practised by every
technologist in the facility. The cleaning procedures in cases of more severe
contamination should always be supervised by the RPO. Local rules should be
110
established that define serious contamination based on radionuclide, activity

and whether it is contamination of a person or equipment and facilities. It is
recommended that the facility have an emergency kit readily available in case of
contamination. Such a kit should contain:
Protective clothing, e.g. overshoes, gloves;
Decontamination materials for the affected areas, including absorbent
materials for wiping up spills;
Decontamination materials for persons;
Warning notices;
Portable monitoring equipment (in working order and regularly checked);
Bags for waste, tape, labels, pencils.
Several severe accidents in medical exposures in nuclear medicine
have been reported and are solely associated with radionuclide therapy and
especially when using 131I in treatment of thyroid diseases. Several incidents
with misadministration of radiopharmaceuticals in diagnostic nuclear medicine
have also been reported. These include examination of the wrong patient or
administration of the wrong radiopharmaceutical or the wrong activity. The most
common incident is to administer the wrong radiopharmaceutical. Even if this
does not cause severe injury to the patient, it is a non-justified exposure with
increased radiation risks. It will also lead to a delayed diagnosis, increased cost
and increased workload because the examination will have to be repeated. Last
but not least, it will cause reduced confidence in the practice of nuclear medicine.
Other accidents and incidents that also involve the general public include
the possible death of a patient containing radionuclides. In diagnostic nuclear
medicine, such an incident can generally be left without specific measures.
However, in radionuclide therapy, emergency plans have to be available on
how to handle the cadaver. Since this is a sensitive issue, depending on ethical
and religious rules and traditions, advice should be available from the national
authorities.
3.8.3. Reporting and lessons learned
In the event of an incident or accident, the licensee has the responsibility
to ensure that a comprehensive investigation takes place and a report is produced
that includes the following information:
A description of the incident by all persons involved;
111
CHAPTER 3
Methods used to estimate the radiation dose received by those involved

in the incident and implications of those methods for possible subsequent
litigation;
Methods used to analyse the incident and to derive risk estimates from the
data;
The subsequent medical consequences for those exposed;
The particulars of any subsequent legal proceedings that may ensue;
Conclusions drawn from the evaluation of the incident and recommendations
on how to prevent a recurrence of such an accident.
In the case of a misadministration or an accident in radionuclide therapy,
the responsible nuclear medicine specialist should be promptly informed. They
should then inform the referring physician and the patient. The medical physicist
should make dose calculations and the staff involved in the accident should
independently describe their view of the accident. Conclusions regarding any
deficits in the procedures should be drawn and necessary changes implemented.
Finally, the licensee may need to submit the report to the regulatory body.
In order to avoid future accidents, it is important to learn from previous
ones. The initiating event and the contributing factors can always be identified.
This information provides material that should be used to prevent future
accidents. This is achieved by informing all members of staff about the accident
or incident, which means that it is very important to have an efficient reporting
system and a programme for local education and training that also includes
potential exposures.
3.9. QUALITY ASSURANCE
3.9.1. General considerations
The International Organization for Standardization defines QA as all
planned and systematic actions needed to provide confidence that a structure,
system or component will perform satisfactorily in service. Satisfactory
performance in nuclear medicine implies the optimum quality of the entire
process. Since an examination or therapy is justified only if the procedure
benefits the patient, QA in the whole process of nuclear medicine is an important
aspect of radiation protection.
The BSS require the licensee of the nuclear medicine facility to have
established a QA programme that provides adequate assurance that the specified
requirements relating to protection and safety are satisfied, and that provides
112
quality control mechanisms and procedures for reviewing and assessing the
overall effectiveness of protection and safety measures.
It is a common and growing practice that hospitals or clinics implement a
quality management system for all of the medical care received in diagnosis and
treatment, i.e. covering the overall nuclear medicine practice. The QA programme
envisaged by the BSS should be part of the wider facility quality management
system. In the hospital or clinic, it is common to include QA as part of the RPP
or, conversely, to include the RPP as part of a more general QA programme for
the hospital or clinic. Regardless of its organization, it is important that radiation
protection is an integral part of a system of quality management. The remainder
of this section considers aspects of QA applied to a nuclear medicine facility that
are covered in the BSS. Specific details with respect to medical exposure are
covered in Section 3.7.2.9.
An effective QA programme requires a strong commitment from the
nuclear medicine facilitys management to provide the necessary resources of
time, personnel and budget. It is recommended that the nuclear medicine facility
establish a group that actively works with QA issues. Such a QA committee
should have a representative from management, a nuclear medicine physician, a
medical physicist, a nuclear medicine technologist and an engineer as members.
The QA committee should meet regularly and review the different components of
the programme.
The QA programme should cover the entire process from the initial decision
to adopt a particular procedure through to the interpretation and recording
of results, and should include ongoing auditing, both internal and external, as
a systematic control methodology. The maintenance of records is an important
part of QA. One important aspect of any QA programme is continuous quality
improvement. This implies a commitment of the staff to strive for continuous
improvement in the use of unsealed sources in diagnosis and therapy, based
on new information learned from their QA programme and new techniques
developed by the nuclear medicine community at large. Feedback from
operational experience and lessons learned from accidents or near misses can
help identify potential problems and correct deficiencies, and should, therefore,
be used systematically, as part of the continuous quality improvement.
A QA programme should cover all aspects of diagnosis and therapy,
including:
The prescription of the procedure by the medical practitioner and its
documentation (supervising if there is any error or contraindication);
Appointments and patient information;
Clinical dosimetry;
Optimization of examination protocol;
113
CHAPTER 3
Record keeping and report writing;

Quality control of radiopharmaceuticals and radionuclide generators;
Acceptance and commissioning;
Quality control of equipment and software;
Waste management procedures;
Training and continuing education of staff;
Clinical audit;
General outcome of the nuclear medicine service.
Further details on the general components of a QA programme and
the associated quality control tests are given in Ref.[3.3]. The WHO has also
published guidelines on QA in nuclear medicine [3.12], covering the organization
of services, the training of personnel, the selection of procedures, quality control
requirements for instrumentation and radiopharmaceuticals, as well as the
interpretation and evaluation of results. The IAEA has several other relevant
publications on QA for various aspects of nuclear medicine (seethe Bibliography
for details).
3.9.2. Audit
The QA programme should be assessed on a regular basis either as an
external or internal audit or review. Audits of activities within the QA programme
should be scheduled on the basis of the status and importance of the activity.
Management should establish a process for such assessments to identify and
correct administrative and management problems that may prevent achievement
of the objectives. Audits and reviews should be conducted by persons who are
technically competent to evaluate the processes and procedures being assessed,
but do not have any direct responsibility for those activities. These may be staff
from other work areas within the organization (internal audit), or an independent
assessment by other organizations (external audit). External audits are generally
a requirement for an accredited practice.
The quality audit should be performed in accordance with written
procedures and checklists. It should include medical, technical and procedural
checks, with the objective to enhance the effectiveness and efficiency of the QA
programme. Any major changes in the QA programme should initiate an audit.
The result should be documented and necessary correction initiated and followed
up.
114
REFERENCES
[3.1] INTERNATIONAL COMMISSION ON RADIOLOGICAL
Recommendations of the ICRP, Publication 103, Elsevier (2008).
PROTECTION,
[3.2] EUROPEAN COMMISSION, FOOD AND AGRICULTURE ORGANIZATION

OF THE UNITED NATIONS, INTERNATIONAL ATOMIC ENERGY AGENCY,
INTERNATIONAL LABOUR ORGANIZATION, OECD NUCLEAR ENERGY
AGENCY, PAN AMERICAN HEALTH ORGANIZATION, UNITED NATIONS
ENVIRONMENT PROGRAMME, WORLD HEALTH ORGANIZATION, Radiation
Protection and Safety of Radiation Sources: International Basic Safety Standards,
IAEA Safety Standards Series No. GSR Part 3, IAEA, Vienna (2014).
[3.3] INTERNATIONAL ATOMIC ENERGY AGENCY, Applying Radiation Safety
Standards in Nuclear Medicine, Safety Reports Series No. 40, IAEA, Vienna (2005).
[3.4] INTERNATIONAL ATOMIC ENERGY AGENCY, Nuclear Medicine Resources
Manual, IAEA, Vienna (2006).
[3.5] INTERNATIONAL ATOMIC ENERGY AGENCY, Occupational Radiation
Protection, IAEA Safety Standards Series No. RS-G-1.1, IAEA, Vienna (1999).
[3.6] INTERNATIONAL ATOMIC ENERGY AGENCY, Assessment of Occupational
Exposure Due to Intakes of Radionuclides, IAEA Safety Standards Series
No. RS-G-1.2, IAEA, Vienna (1999).
[3.7] INTERNATIONAL ATOMIC ENERGY AGENCY, Assessment of Occupational
Exposure Due to External Sources of Radiation, IAEA Safety Standards Series
No. RS-G-1.3, IAEA, Vienna (1999).
[3.8] INTERNATIONAL ATOMIC ENERGY AGENCY, Management of Waste from
the Use of Radioactive Material in Medicine, Industry, Agriculture, Research and
Education, IAEA Safety Standards Series No. WS-G-2.7, IAEA, Vienna (2005).
[3.9] INTERNATIONAL ATOMIC ENERGY AGENCY, Regulations for the Safe Transport
of Radioactive Material, 2012 Edition, IAEA Safety Standards Series No. SSR-6,
IAEA, Vienna (2012).
[3.10] WORLD MEDICAL ASSOCIATION, 18th World Medical Assembly, Helsinki, 1974,
as amended by the 59th World Medical Assembly, Seoul (2008).
[3.11] COUNCIL FOR INTERNATIONAL ORGANIZATIONS OF MEDICAL SCIENCES,
WORLD HEALTH ORGANIZATION, International Ethical Guidelines for
Biomedical Research Involving Human Subjects, CIOMS, Geneva (2002).
[3.12] WORLD HEALTH ORGANIZATION, Quality Assurance in Nuclear Medicine,
WHO, Geneva (1982).
115
CHAPTER 3
BIBLIOGRAPHY
EUROPEAN COMMISSION, European Guidelines on Quality Criteria for Computed
Tomography, Rep. EUR 16262 EN, Office for Official Publications of the European
Communities, Brussels (1999).
INTERNATIONAL ATOMIC ENERGY AGENCY (IAEA, Vienna)
Quality Control of Nuclear Medicine Instruments 1991, IAEA-TECDOC-602 (1991).
Radiological Protection for Medical Exposure to Ionizing Radiation, IAEA Safety Standards
Series No. RS-G-1.5 (2002).
IAEA Quality Control Atlas for Scintillation Camera Systems (2003).
Quality Assurance for Radioactivity Measurement in Nuclear Medicine, Technical Reports
Series No. 454 (2006).
Radiation Protection in Newer Medical Imaging Techniques: PET/CT, Safety Reports Series
No. 58 (2008).
Quality Assurance for PET and PET/CT Systems, IAEA Human Health Series No. 1 (2009).
Quality Assurance for SPECT Systems, IAEA Human Health Series No. 6 (2009).
Quality Management Audits in Nuclear Medicine Practices (2009).
Radiation Protection of Patients (RPoP),
https://rpop.iaea/RPOP/RPOP/Content/index.htm
INTERNATIONAL COMMISSION ON RADIOLOGICAL PROTECTION
Radiological Protection of the Worker in Medicine and Dentistry, Publication 57, Pergamon
Press, Oxford and New York (1989).
Radiological Protection in Biomedical Research, Publication 62, Pergamon Press, Oxford and
New York (1991).
Radiological Protection in Medicine, Publication 105, Elsevier (2008).
Pregnancy and Medical Radiation, Publication 84, Pergamon Press, Oxford and New York
(2000).
INTERNATIONAL COMMISSION ON RADIATION UNITS AND MEASUREMENTS,
Quantities and Units in Radiation Protection Dosimetry, ICRU Rep. 51, Bethesda MD (1993).
MADSEN, M.T., et al., AAPM Task Group 108: PET and PET/CT shielding requirements,
Med. Phys. 33 (2006) 1.
SMITH, A.H., HART, G.C. (Eds), INSTITUTE OF PHYSICAL SCIENCES IN MEDICINE,
Quality Standards in Nuclear Medicine, IPSM Rep. No. 65, York (1992).
116
CHAPTER 4
RADIONUCLIDE PRODUCTION
H.O. LUNDQVIST
Department of Radiology, Oncology and Radiation Science,
Uppsala University,
Uppsala, Sweden
4.1. THE ORIGINS OF DIFFERENT NUCLEI
All matter in the universe has its origin in an event called the big bang,
a cosmic explosion releasing an enormous amount of energy about 14 billion
years ago. Scientists believe that particles such as protons and neutrons, which
form the building blocks of nuclei, were condensed as free particles during the
first seconds. With the decreasing temperature of the expanding universe, the
formation of particle combinations such as deuterium (heavy hydrogen) and
helium occurred. For several hundred million years, the universe was plasma
composed of hydrogen, deuterium, helium ions and free electrons. As the
temperature continued to decrease, the electrons were able to attach to ions,
forming neutral atoms and converting the plasma into a large cloud of hydrogen
and helium gas. Locally, this neutral gas slowly condensed under the force of
gravity to form the first stars. As the temperature and the density in the stars
increased, the probability of nuclear fusion resulting in the production of heavier
elements increased, culminating in all of the elements in the periodic table that
we know today. As the stars aged, consuming their hydrogen fuel, they eventually
exploded, spreading their contents of heavy materials around the universe.
Owing to gravity, other stars formed with planets around them, composed of
these heavy elements. Four and a half billion years have passed since the planet
Earth was formed. In that time, most of the atomic nuclei consisting of unstable
protonneutron combinations have undergone transformation (radioactive
decay) to more stable (non-radioactive) combinations. However, some with very
long half-lives remain: 40K, 204Pb, 232Th and the naturally occurring isotopes of
uranium.
The discovery of these radioactive atoms was first made by Henri Becquerel
in 1896. The chemical purification and elucidation of some of the properties
of radioactive substances was further investigated by Marie Skodowska-Curie
and her husband Pierre Curie. Since some of these long lived radionuclides
generated more short lived radionuclides, a new scientific tool had been
117
CHAPTER 4
discovered that was later found to have profound implications in what today is
known as nuclear medicine. George de Hevesy was a pioneer in demonstrating
the practical uses of the new radioactive elements. He and his colleagues used a
radioactive isotope of lead, 210Pb, as a tracer (or indicator) when they studied the
solubility of sparingly soluble lead salts. De Hevesy was also the first to apply
the radioactive tracer technique in biology when he investigated lead uptake in
plants (1923) using 212Pb. Only one year later, Blumengarten and Weiss carried
out the first clinical study, when they injected 212Bi into one arm of a patient and
measured the arrival time in the other arm. From this study, they concluded that
the arrival time was prolonged in patients with heart disease.
4.1.1. Induced radioactivity
In the beginning, nature was the supplier of the radioactive nuclides
used. Isotopes of uranium and thorium generated a variety of radioactive
heavy elements such as lead, but radioactive isotopes of light elements were
not known. Marie Curies daughter Irne, together with her husband Frdric
Joliot took the next step. Alpha emitting sources had long been used to bombard
different elements, for example, by Ernest Rutherford who studied the deflection
of particles in gold foils. The large deflections observed in this work led to
the conclusion that the atom consisted of a tiny nucleus of protons with orbiting
electrons (similar to planets around the sun). However, JoliotCurie also showed
that the particles induced radioactivity in the bombarded foil (in their case,
aluminium foil). The induced radioactivity had a half-life of about 3min. They
identified the emitted radiation to be from 30P created in the nuclear reaction
27
Al(, n)30P.
They also concluded that:
These elements and similar ones may possibly be formed in different
nuclear reactions with other bombarding particles: protons, deuterons
and neutrons. For example, 13N could perhaps be formed by capture of a
deuteron in 12C, followed by the emission of a neutron.
The same year, this was proved to be true by Ernest Lawrence in Berkeley,
California and Enrico Fermi in Rome. Lawrence had built a cyclotron capable
of accelerating deuterons up to about 3MeV. He soon reported the production
of 13N with a half-life of 10min. Thereafter, the cyclotron was used to produce
several other biologically important radionuclides such as 11C, 32P and 22Na.
Fermi realized that the neutron was advantageous for radionuclide production.
Since it has no charge, it could easily enter into the nucleus and induce a nuclear
reaction. He immediately made a strong neutron source by sealing up 232Ra gas
118
with beryllium powder in a glass vial. The particle emitted from 232Ra caused a
nuclear reaction in beryllium and a neutron was emitted, 9Be(, n)12C.
Fermi and his research group started a systematic search by irradiating all
available elements in the periodic system with fast and slow neutrons to study
the creation of induced radioactivity. From hydrogen to oxygen, no radioactivity
was observed in their targets, but in the ninth element, fluorine, their hopes were
fulfilled. In the following weeks, they bombarded some 60 elements and found
induced radioactivity in 40 of them. They also observed that the lighter elements
were usually transmuted into radionuclides of a different chemical element,
whereas heavier elements appeared to yield radioisotopes of the same element
as the target.
These new discoveries excited the scientific community. From having been
a rather limited technique, the radioactivity tracer principle could suddenly be
applied in a variety of fields, especially in life sciences. De Hevesy immediately
started to study the uptake and elimination of 32P phosphate in various tissues of
rats and demonstrated, for the first time, the kinetics of vital elements in living
creatures. Iodine-128 was soon after applied in the diagnosis of thyroid disease.
This was the start of the radiotracer technology in biology and medicine as
we know it today.
One early cyclotron produced nuclide of special importance was 11C
since carbon is fundamental in life sciences. Carbon-11 had a half-life of only
20min but by setting up a chemical laboratory close to the cyclotron, organic
compounds labelled with 11C were obtained in large amounts. Photosynthesis
was studied using 11CO2 and the fixation of carbon monoxide in humans by
inhaling 11CO. However, 20min was a short half-life and the use of 11C was
limited to the most rapid biochemical reactions. It must be remembered that the
radio-detectors used at that time were primitive and that the chemical, synthetic
and analytical tools were not adapted to such short times. A search to find a more
long lived isotope of carbon resulted in the discovery in 1939 of 14C produced in
the nuclear reaction 13C(d, p)14C.
Unfortunately, 14C produced this way was of limited use since the
radionuclide could not be separated from the target. However, during the
bombardments, a bottle of ammonium nitrate solution had been standing close
to the target. By pure chance, it was discovered that this bottle also contained
14
C, which had been produced in the reaction 14N(n, p)14C.
The deuterons used in the bombardment consist of one proton and one
neutron with a binding energy of about 2MeV. When high energy deuterons
hit a target, it is likely that the binding between the particles breaks and that a
free neutron is created in what is called a stripping reaction. The bottle with
ammonium nitrate had, thus, unintentionally been neutron irradiated. Since no
carbon was present in the bottle (except small amounts from solved airborne
119
CHAPTER 4
carbon dioxide), the 14C produced this way was of high specific radioactivity.
It was also very easy to separate from the target. In the nuclear reaction, a
hot carbon atom was created, which formed 14CO2 in the solution. By simply
bubbling air through the bottle, the 14C was released from the target.
The same year, tritium was discovered by deuteron irradiation of water.
One of the pioneers Martin Kamen stated:
Within a few months, after the scientific world had somewhat ruefully
concluded that development of tracer techniques would be seriously
handicapped because useful radioactive tracers for carbon, hydrogen,
oxygen and nitrogen did not exist, 14C and 3H were discovered.
Before the second world war, the cyclotron was the main producer of
radionuclides since the neutron sources at that time were very weak. However,
with the development of the nuclear reactor, that situation changed. Suddenly,
a strong neutron source was available, which could easily produce almost
unlimited amounts of radioactive nuclides including biologically important
elements, such as 3H, 14C, 32P and 35S, and clinically interesting radionuclides,
such as 60Co (for external radiotherapy) and 131I, for nuclear medicine. After
the war, a new industry was born which could deliver a variety of radiolabelled
compounds for research and clinical use at a reasonable price.
However, accelerator produced nuclides have a special character, which
makes them differ from reactor produced nuclides. Today, their popularity is
increasing again. Generally, reactor produced radionuclides are most suitable for
laboratory work, whereas accelerator produced radionuclides are more useful
clinically. Some of the most used radionuclides in nuclear medicine, such as
111
In, 123I and 201Tl, and the short lived radionuclides, 11C, 13N, 15O and 18F, used
for positron emission tomography (PET), are all cyclotron produced.
4.1.2. Nuclide chart and line of nuclear stability
During the late 19th century, chemists learned to organize chemical
knowledge into the periodic system. Radioactivity, when it was discovered,
conflicted with that system. Suddenly, various samples, apparently with the
same chemical behaviour, were found to have different physical qualities such
as half-life, emitted type of radiation and energy. The concept of isotopes or
elements occupying the same place in the periodic system (from the Greek
(isos topos) meaning same place) was introduced by Soddy
1913, but a complete explanation had to await the discovery of the neutron by
Chadwick in 1932.
120
The periodic system was organized according to the number of protons

(atom number) in the nucleus, which is equal to the number of electrons to
balance the atomic charge. The nuclide chart consists of a plot with the number
of neutrons in the nucleus on the x axis and the number of protons on the y axis
(Fig.4.1).
Number of protons
NUCLIDE CHART
Number of neutrons
FIG.4.1. Chart of nuclides. The black dots represent 279 naturally existing combinations of
protons and neutrons (stable or almost stable nuclides). There are about 2300 proton/neutron
combinations that are unstable around this stable line.
Figure4.2 shows a limited part of the nuclide chart. The formal notation
for an isotope is ZA X , where X is the element name (e.g. C for carbon), A is the
mass number (A=Z + N), Z is the number of protons in the nucleus (atom
number) and N the number of neutrons in the nucleus.
The expression above is overdetermined. If the element name X is known,
so is the number of protons in the nucleus, Z. Therefore, the simplified notation
A
X is commonly used.
Some relations of the numbers of protons and neutrons have special names
such as:
Isotopes: the number of protons is constant (Z=constant).
Isotones: the number of neutrons is constant (N=constant).
Isobars: The mass number is constant (A=constant).
121
CHAPTER 4
Of these expressions, only the isotope concept is generally used. It is

important to understand that whenever the expression isotope is used, it must
always be related to a specific element or group of elements, for example,
isotopes of carbon (e.g. 11C, 12C, 13C and 14C).
Number of protons
17F
18F
19F
20F
21F
22F
23F
24F
25F
13O
14O
15O
16O
17O
18O
19O
20O
21O
22O
23O
24O
12N
13N
14N
15N
16N
17N
18N
19N
20N
21N
22N
23N
11C
12C
13C
14C
15C
16C
17C
18
19C
20C
10B
11B
12B
13B
14B
15B
11Be
12Be
13
14
15
16
9C
8B
7Be
8Be
9Be
10Be
6Li
7Li
8Li
9Li
2
1
1H
3He
4He
2H
3H
10C
6He
17B
14Be
11Li
8He
n
0
10
11
12
Number of neutrons
FIG.4.2. A part of the nuclide chart where the lightest elements are shown. The darkened
fields represent stable nuclei. Nuclides to the left of the stable ones are radionuclides deficient
in neutrons and those to the right, rich in neutrons.
In the nuclide chart (Fig.4.1), the stable nuclides fall along a monotonically
increasing line called the stability line. The stability of the nucleus is determined
by competing forces: the strong force that binds the nucleons (protons and
neutrons) together and the Coulomb force that repulses particles of like charge,
e.g. protons. The interplay between the forces is illustrated in Fig.4.3.
For best stability, the nucleus has an equal number of protons and
neutrons. This is a quantum mechanic feature of bound particles and in Fig.4.1
this is illustrated by a straight line. It is also seen that the stability line follows
the straight line for the light elements but that there is considerable deviation
(neutron excess) for the heavier elements. The explanation is the large Coulomb
force in the heavy elements which have many protons in close proximity. By
diluting the charge by non-charged neutrons, the distance between the charges
increases and the Coulomb force decreases.
122
FIG.4.3. Between the proton and a neutron, there is a nuclear force that amounts to
2.225MeV. The nucleons form a stable combination called deuterium, an isotope of hydrogen.
In a system of two protons, the nuclear force is equally strong to a neutronproton, but the
repulsive Coulomb forces are stronger. Thus, this system cannot exist. The nuclear force
between two neutrons is equally strong and there is no Coulomb force. Nevertheless, this
system cannot exist due to other repulsive forces, a consequence of the rules of pairing quarks.
4.1.3. Binding energy, Q-value, reaction threshold

and nuclear reaction formalism
There are no barriers and no repulsive forces between a free proton and
neutron, and they can fuse at low kinetic energies to form a deuterium nucleus,
which has a weight somewhat smaller than the sum of the free neutron and
proton weights. This mass difference can be converted into energy using
Albert Einsteins formula E=mc2 and is found to be 2.2MeV. This is also the
energy released as a photon in the reaction. To separate the two nucleons in
the deuterium nucleus, at least 2.2MeV have to be added. The energy gained
or lost in a nuclear reaction is called the Q-value. In a somewhat more complex
reaction, 14N(p, )11C, the Q-value is calculated as the difference between the
summation of the mass of the particles before the reaction (p, 14N) from the mass
of the particles after the reaction (, 11C). It should be noted that it is the mass
of the nucleus and not the atomic mass that is used. Using a Q-value calculator1,
the Q-value for the reaction 14N(p, )11C is 2923.056keV. This means that the
proton, when it reaches the 14N nucleus, has to have a kinetic energy of at least
2.93MeV in order to make the reaction possible.
However, before it hits the nucleus, the proton has to overcome the barrier
created by the repulsive Coulomb force between the proton and the positive 14N
nucleus. During the passage, the proton loses some energy and the starting value,
called the threshold value, must then exceed the Q-value. The same calculator
gives the threshold value of 3.14MeV for the 11C production reaction.
The reaction energy (the Q-value) is positive for exothermal reactions
(spontaneous reactions) and negative for endothermal reactions. Since all radioactive
decays are spontaneous, they need to have positive Q-values. Some reactions used
For example, http://nucleardata.nuclear.lu.se/database/masses/
123
CHAPTER 4
to produce radionuclides, mainly those that are based upon thermal neutrons, have
positive Q-values but reactions based on positive particles usually have negative
Q-values, e.g. extra energy needs to be added to get the reaction going.
4.1.4. Types of nuclear reaction, reaction channels and cross-section
As seen in Fig.4.1, the radionuclides to the right of the stability line have
an excess of neutrons compared to the stable elements and they are preferentially
produced by irradiating a stable nuclide with neutrons. The radionuclides to
the left are neutron deficient or have an excess of charge and, hence, they are
mainly produced by irradiating stable elements by a charged particle, e.g. p or d.
Although these are the main principles, there are exceptions.
Usually, the irradiating particles have a large kinetic energy that is
transferred to the target nucleus to enable a nuclear reaction (the exception
being thermal neutrons that can start a reaction by thermal diffusion). Figure4.4
shows schematically an incoming beam incident upon the target, where it may
be scattered and absorbed. It can transfer its energy totally or partly to the target
nucleus and can interact with parts of or the whole of the target nucleus. Since the
produced activity should be high, the target is also usually thick.
d
I0Nd d/d
I0
Incident beam
Target
Scattered particles
FIG.4.4. Target irradiation. A nuclear physicist is usually interested in the particles coming
out, their energy and angular distribution, but the radiochemist is mainly interested in the
transformed nuclides in the target.
In radionuclide production, the nuclear reaction always involves a change

in the number of protons or neutrons. Reactions that result in a change in the
124
number of protons are preferable because the product becomes a different

element, facilitating chemical separation from the target, compared to an (n, )
reaction, where the product and target are the same.
Neutrons can penetrate the target at down to thermal energies. Charged
particles need to overcome the Coulomb barrier to penetrate the nucleus (Fig.4.5).
FIG.4.5. General cross-sectional behaviour for nuclear reactions as a function of the incident
particle energy. Since the proton has to overcome the Coulomb barrier, there is a threshold that
is not present for the neutron. Even very low energy neutrons can penetrate into the nucleus to
cause a nuclear reaction.
The parameter cross-section is the probability of a certain nuclear reaction

happening and is expressed as a surface. It is the probability that a particle will
interact per unit surface area of target. The geometrical cross-section of a uranium
nucleus is roughly 1028 m2, and this area has also been taken to define the unit
for cross-section barn (b). This is not an International System of Units unit but is
commonly used to describe reaction probabilities in atomic and nuclear physics.
For fast particle reactions, the probability is usually less than the
geometrical cross-section area of the nucleus, with probabilities in the range
of millibarns. However, the probability of a hit is a combination of the area of
both the nucleus and the incoming particle. The Heisenberg uncertainty principle
states that the position and the momentum of particles cannot be simultaneously
known to arbitrarily high precision. This implies that particles of well defined
but low energy, such as thermal neutrons, will have a large uncertainty in their
position. One may also say that they are increasing in size and nuclear reactions
involving thermal neutrons may have very large cross-sections, sometimes of the
order of several thousand barns.
125
CHAPTER 4
The general equation for a nuclear reaction is:

a+Ab+B+Q
where a is the incoming particle and A is the target nucleus in the ground state (the
entrance channel). Depending on the energy and the particle involved, several
nuclear reactions may happen, each with its own probability (cross-section). Each
nuclear reaction creates an outgoing channel, where b is the outgoing particle
or particles and B is the rest nucleus. Q is the reaction energy and can be both
negative and positive.
A common notation of a nuclear reaction is A(a, b)B. If the incoming
particle is absorbed, there is a capture process type (n, ) and in a reaction of type
(p, n) charge exchange occurs. If many particles are expelled, the reaction can
be referred to as (p, 3n). Each such reaction is called a reaction channel and is
characterized by an energy threshold (an energy that makes the nuclear reaction
possible, opens up the channel) and a probability (cross-section) varying with the
incoming particle energy. A schematic illustration of different reaction channels
opened in proton irradiation is given in Fig.4.6.
FIG.4.6. A schematic figure showing some reaction channels upon proton irradiation.
Different reaction mechanisms can operate in the same reaction channel.

Here, two ways are differentiated:
The formation of a compound nucleus;
Direct reactions.
126
The compound nucleus has a large probability to be formed in a central

hit of the nucleus and is preferable at low energies close to the energy threshold
of the reaction channel. Here, the incoming particle is absorbed and an excited
compound nucleus formed. This compound nucleus will rapidly (~1019 s)
undergo decay (fragment) with the isotropic emission of neutrons and rays.
Direct reactions preferentially occur at the edge of the nucleus or at high energies.
The incoming energy is directly transferred to a nucleon (knock-on reaction)
giving two outgoing particles. The outgoing particles usually have high energy
and are emitted in about the same direction as the incoming particle.
FIG.4.7. A schematic view of particle energy variations of a cross-section for direct nuclear
reactions and for forming a compound nucleus.
The production of radionuclides is due to a mixture of these two reaction

types. Their probability varies with energy in different ways. The direct
reactions are heavily associated with the geometrical size of the nucleus, and
the cross-section is usually small and fairly constant with energy. The highest
probability of forming a compound nucleus is just above the reaction threshold as
seen in Fig.4.7.
4.2. REACTOR PRODUCTION
There are two major ways to produce radionuclides: using reactors
(neutrons) or particle accelerators (protons, deuterons, particles or heavy ions).
Since the target is a stable nuclide, either a neutron-rich radionuclide (reactor
127
CHAPTER 4
produced) or a neutron deficient radionuclide (accelerator produced) is generally

obtained.
4.2.1. Principle of operation and neutron spectrum
A nuclear reactor is a facility in which a fissile atomic nucleus such as 235U,
Pu or 241P absorbs a low energy neutron and undergoes nuclear fission. In the
process, fast neutrons are produced with energies from about 10MeV and below
(the fission neutron spectrum). The neutrons are slowed down in a moderator
(usually water) and the slowed down neutrons start new fissions. By regulating
this nuclear chain reaction, there will be a steady state of neutron production with
a typical neutron flux of the order of 1014 neutrons cm2 s1.
Since neutrons have no charge and are, thus, unaffected by the Coulomb
barrier, even thermal neutrons (0.025 eV) can enter the target nucleus and
cause a nuclear reaction. However, some nuclear reactions, depending upon the
cross-section, require fast neutrons (energy < 10MeV).
A reactor produces a neutron cloud in which the target is placed so that it
will be isotropically irradiated. Placing the target in different positions exposes
it to neutrons of different energy. Usually, the reactor facility has a pneumatic
system for placing targets at predefined positions. One has to consider the heat
that is generated in the reactor core, since the temperature at some irradiation
positions may easily reach 200C. The reactor is characterized by the energy
spectrum, the flux (neutrons cm2 s1) and the temperature at the irradiation
position.
Most reactors in the world are for energy production and, for safety reasons,
cannot be used for radionuclide production. Usually, only national research
reactors are flexible enough for use in radioisotope production.
239
4.2.2. Thermal and fast neutron reactions

The most typical neutron reaction is the (n, ) reaction in which a thermal
neutron is captured by the target nucleus forming a compound nucleus. The
decay energy is emitted as a prompt ray. A typical example is the reaction
59
Co(n, )60Co that produces an important radionuclide used in external therapy.
However, since the produced radionuclide is of the same element as the target, the
specific activity a, i.e. the radioactivity per mass of the sample, is low. This type
of nuclear reaction is of little interest when labelling radiopharmaceuticals. In
light elements, other nuclear reactions resulting from thermal neutron irradiation
are possible, such as (n, p). Table4.1 lists possible production reactions for some
biologically important radionuclides.
128
TABLE4.1. TYPICAL NUCLEAR REACTIONS IN A REACTOR FOR

Type of neutrons
Nuclear reaction
Thermal
59
60
Co(n, ) Co
14
N(n, p)14C
33
33
S(n, p) P
35
Cl(n, )32P
Fast
Half-life T1/2
Cross-section (mb)
5.3 a
2000
5730 a
1.75
25 d
0.015
24 d
0.05
24
15 h
1.2
Cl(n, )32P
14 d
6.1
24
Mg(n, p) Na
35
Nuclear reactions with thermal neutrons are attractive for many reasons.
The yields are high due to large cross-sections and the high thermal neutron fluxes
available in the reactor. In some cases, the yields are sufficiently high to use these
reactions as the source of charged secondary particles, e.g. 6Li(n, )3H for the
production of high energy 3H ions, which can then be used for the production of
18
F by 16O(3H, n)18F. The target used is 6LiOH, in which the produced 3H ions
will be in close contact with the target 16O. A drawback of this production is that
when the target is dissolved the solution is heavily contaminated with 3H water
that might be difficult to remove. Today, with an increasing number of hospital
based accelerators, there is little need of neutron produced 18F.
Another reactor produced neutron deficient radionuclide is 125I:
124
Xe(n, )125Xe (T1/2=17 h) 125I (T1/2=60 d)
This is currently the common way of producing high quality 125I. A

drawback is that 124Xe has a natural abundance of 0.1%. To increase the
production yield, one needs to work with expensive enriched targets. However,
these can be reused several times. This is an example of a generator system where
the mother is shorter lived than the daughter. Although there is no need to make a
separation between the mother and daughter, the target, after irradiation, has to be
stored for some days to allow the decay of 125Xe to be complete. The expensive
target gas 124Xe is carefully removed and the 125I is washed out from the walls of
the target capsule.
Many reactor produced radionuclides emit high energy particles that
contribute to the absorbed dose (but not the imaging signal) to patients, which is
a drawback in diagnostic procedures. However, a few emitting isotopes result
in daughter nuclei that emit rays with long de-excitation times (metastable
129
CHAPTER 4
excited levels), instead of the more common prompt (1014 s) emission. Such
radioisotopes are suitable for nuclear medicine imaging, since they principally
yield radiation, with some electron emission, a consequence of internal
conversion. The most commonly used radionuclide in nuclear medicine, 99mTc,
is of this type. The m after the atomic mass signifies that this is the metastable
version of the radionuclide.
In radionuclide therapy, in contrast to diagnostic applications, the emission
of high energy radiation is desirable. Most radionuclides for radiotherapy are,
therefore, reactor produced. Examples include 90Y, 131I and 177Lu. A case of
interest to study is 177Lu, which can be produced in two different ways using
thermal neutrons. The most common production route is still the (n, ) reaction
on 176Lu, which opposes two conventional wisdoms in practical radionuclide
production for biomolecular labelling:
(a) Not to use a production that yields the same product element as the target
since it will negatively affect the labelling ability due to the low specific
radioactivity;
(b) Not to use a target that is radioactive.
However, 176Lu is a natural radioactive isotope of lutetium with an
abundance of 2.59%. Figure4.8 shows how 177Lu needs to be separated from
the dominant 175Lu to decrease the mass of the final product. This method of
production works because the high cross-section (2020 b) of 176Lu results in a
high fraction of the target atoms being converted to 177Lu, yielding an acceptable
specific radioactivity of the final product.
On the right of Fig.4.8, an indirect way to produce 177Lu from 176Yb is also
shown. This method of production utilizes a generator nuclide 177Yb, produced
by an (n, ) reaction, which then decays to 177Lu. In principle, by chemically
separating lutetium from ytterbium, one would obtain the highest possible specific
radioactivity. However, the chemical separation between two lanthanides is not
trivial and, thus, it is difficult to obtain 177Lu without substantial contamination of
the target material Yb that may compete in the labelling procedure. Furthermore,
the cross-section for this reaction is almost a thousandfold lower, resulting in a
much lower product yield.
Reactions involving fast neutrons usually have cross-sections that are of the
order of millibarns, which, coupled with the much lower neutron flux at higher
energy relative to thermal neutron fluxes, leads to lower yields. However, there
are some important radionuclides, e.g. 32P that have to be produced this way.
Figure4.9 gives the details of this production.
130
T1/2
Abundance (%)
(mb)
175Lu
Stable
97.41
176Lu
3.78 1010 a
2.59
2020
177Lu
6.734 d
178Lu
28.4 min
175Lu
Stable
97.41
174Yb
Stable
31.8
T1/2
Abundance (%)
(mb)
176Lu
3.78 1010 a
2.59
2020
175Yb
4.185 d
177Lu
6.734 d
178Lu
28.4 min
176Yb
Stable
12.7
2.85
177Yb
1.911 h
31.8
FIG. 4.8. Production of 177Lu from 176Lu (left) and from 176Yb (right).
32S
Stable
95.02
varying
Cross-section31P
T1/2
Stable
Abundance (%)
100
(mb)
T1/2
Abundance (%)
(mb)
33S
Stable
0.75
varying
32
32P S
14.26 d
34S
Stable
4.21
varying
(n,p)3332PP
25.4 d
T1/2
Abundance (%)
(mb)
T1/2
Abundance (%)
(mb)
32S
Stable
95.02
varying
31P
Stable
100
33S
Stable
0.75
varying
32P
14.26 d
34S
Stable
4.21
varying
33P
25.4 d
Neutron energy
FIG. 4.9. Data for the production of 32P in the nuclear reaction 32S(n, p)32P. The reaction
threshold is 0.51 MeV. From the cross-section data, it can be seen that there is no substantial
yield until an energy of about 2 MeV. The yield is an integration of the cross-section data and
the neutron energy spectrum. A practical cross-section can be calculated to about 60 mb.
4.2.3. Nuclear fission, fission products

Uranium-235 is not only used as fuel in a nuclear reactor but it can also be
used as a target to produce radionuclides. Uranium-235 irradiated with thermal
neutrons undergoes fission with a cross-section of 586 b. The fission process
results in the production of two fragments of 235U nucleus plus a number of free
neutrons. The sum of the fragments mass will be close to the mass of 235U, but
they will vary according to Fig. 4.10.
The masses of the 99Mo and 134Sn produced in the reaction:
235
U + n 236U 99Mo + 134Sn + 3n
are marked in Fig. 4.10. Some medically important radionuclides are produced
by fission, such as 90Y (therapy) and 99mTc (diagnostic). They are not produced
directly but by a generator system:
90
Sr (28.5 a) 90Y (2.3 d) and 99Mo (2.7 d) 99mTc (6 h)
131
CHAPTER 4
U-235, fission yield
Yield/mass number (%)
Mass 134
Mass 99
Mass number
FIG.4.10. The yield of fission fragments as a function of mass.
The primary radionuclides produced are then 90Sr and 99Mo, or more
precisely the mass numbers 90 and 99.
Another important fission produced radionuclide in nuclear medicine,
both for diagnostics and therapy, is 131I. The practical fission cross-section for
this production is the fission cross-section of 235U multiplied by the fraction
of fragments having a mass of 131 or 5860.029=17 b. The probability of
producing a mass of 131 is 2.9% per fission. Iodine-131 is the only radionuclide
with a mass of 131 that has a half-life of more than 1 h, meaning that all of the
others will soon have decayed to 131I.
4.3. ACCELERATOR PRODUCTION
Charged particles, unlike neutrons, are unable to diffuse into the nucleus,
but need to have sufficient kinetic energy to overcome the Coulomb barrier.
However, charged particles are readily accelerated to kinetic energies that open
up more reaction channels than fast neutrons in a reactor. An example is seen in
Fig.4.11 that also illustrates alternative opportunities with: p, d, 3He and 4He or
, to produce practical and economic nuclear reactions.
132
127
i(p, 5n) 123 Xe 123 i
124
Xe(p, np) 123 Xe 123 i
123
Te(p, n) 123 i
122
Te(d, n) 123 i
124
Te(p, 2n) 123 i
121
sb( 4 he, 2n) 123 i
121
sb( 3 he, n) 123 i
123
sb( 3 he, 3n) 123 i
FIG.4.11. Various nuclear reactions that produce 123I. All of the reactions have been tried
and can be performed at relatively low particle energies. The 123Xe produced in the first two
reactions decays to 123I with a half-life of about 2 h. In the first reaction, the 123Xe is separated
from the target and then decays, while in the second reaction, the 123I is washed out of the
target after decay.
An accelerator in particle physics can be huge, as in the European

Organization for Nuclear Research (CERN), with a diameter of more than 4 km.
Accelerators for radionuclide production are much smaller, as they need to
accelerate particles to much lower energies. The first reaction in Fig.4.11, where
five neutrons are expelled, is the most energy demanding, as it requires a proton
energy of about 510MeV=50MeV (the rule of thumb is that about 10MeV
are required per expelled particle). All of the other reactions require 20MeV or
less (seeTable4.2).
Another advantage with accelerator production is that it is usually easy to
find a nuclear reaction where the product is a different element from the target.
Since different elements can be separated chemically, the product can usually be
of high specific radioactivity, which is important when labelling biomolecules.
A technical difference between reactor and accelerator irradiation is
that in the reactor the particles come from all directions but in the accelerator
the particles have a particular direction. The number of charged particles is
often smaller and is usually measured as an electric current in microamperes
(1 A=61012 protons/s but 31012 alpha/s because of the two charges of the
particle).
133
CHAPTER 4
TABLE4.2. CHARACTeRIZATION
Proton energy
(MeV)
OF
ACCELERATORS
Accelerated particles
Used for
<10
Mainly single particle, p or d
PET
1020
Usually p and d
PET
FOR
3040
p and d, He and He may be available
PET, commercial production
40500
Usually p only
Often placed in national

centres and have several
users
A drawback in accelerator production is that charged particles are stopped

more efficiently than neutrons; for example, 16MeV protons are stopped in
0.6mm Cu. A typical production beam current of 100 A hitting a typical target
area of 2cm2 will then put 1.6 kW in a volume of 0.1cm3, which will evaporate
most materials if not efficiently cooled. In addition, the acceleration of the beam
occurs in a vacuum but the target irradiation is at atmospheric pressure or in gas
targets at 1020 times over pressure. To separate the vacuum from the target, the
beam has to penetrate foils that will absorb some particle energy and they will
also become strongly activated.
4.3.1. Cyclotron, principle of operation, negativeandpositiveions
There are several types of accelerator, all of which can, in principle, be
used for radionuclide production. The dominant one for radionuclide production
is currently the cyclotron that was invented by Lawrence in the early 1930s.
Cyclotrons were first installed in hospitals in the 1960s, but during the past two
decades, hospital based small cyclotrons yielding 1020MeV protons have
become fairly common, especially with the rise of PET.
A cyclotron is composed of four systems:
(a) A resistive magnet that can create a magnetic field of 12 T;
(b) A vacuum system down to 105 Pa;
(c) A high frequency system (about 40 MHz) providing a voltage with a peak
value of about 40kV, although these figures can vary considerably for
different systems;
(d) An ion source that can ionize hydrogen to create free protons as well as
deuterium and particles.
134
FIG.4.12. The cyclotron principle. A negative ion is injected into the gap between D-shaped
magnets (Dees) (1). An alternating electric field is applied across the gap, which causes
FIG.
4.12.
The cyclotron
principle.force
A negative
ion charge
is injected
dee
the
charge
to accelerate.
The magnetic
on a moving
forcesinto
it to the
bendgap
intobetween
a
shaped magnets
Anincreasing
alternating
electric
fieldapplied
is applied
across
which
semicircular
path of(1).
ever
radius
(2). The
electric
fieldthe
is gap,
reversed
in causes the
chargeeach
to accelerate.
Theparticle
magnetic
force
on aso moving
charge forces
it to bend into a
direction
time the charged
reaches
the gap,
that it is continuously
accelerated,
semicircular
of (3).
ever increasing radius (2). The applied electric field is reversed in
until
finally beingpath
ejected
direction each time the charged particle reaches the gap, so that it is continuously
accelerated, until finally being ejected (3).
The inside of a cyclotron is shown in Fig.4.12. The ion source is usually

placed
the vacuum
and in
thehigh
centre
(internal),
in larger
machines,it is usual to
In inside
commercial
accelerators,
with
beam
currents but,
of several
milliamperes,
can
beanexternal.
then injected
from located
the outside
a central
have
internal The
targetions
for are
radionuclide
production
insidethrough
the chamber.
In accelerators
hole
the magnet.
The main
idea
of such
the ion
is to havefor
a slow
of gas
withinlower
beam currents
<100
A,
as source
those dedicated
PET flow
hospital
facilities, it is
more
common
extract
onto an The
external
target
modes of extraction
that
is made
into to
plasma
bythe
an beam
arc discharge.
desired
ionsystem.
species The
are extracted
depend aupon
whetherand
positive
or negative
ions are
accelerated.
Extraction
of positive ions is
through
collimator
accelerated
in a static
electric
field. There
are several
made
by
using
a
deflector
that
applies
a
static
electric
field
which
acts
upon
the particles when
types of ion source with different operating characteristics. In modern accelerators,
in the outer orbits. Some beam current is invariably lost in the process and the deflector often
negative ions, protons or deuterium with two orbit electrons are usually used. These
becomes quite radioactive.
facilitate
extraction
of the beam. accelerators usually accelerate negative ions that are more
Modern
proton/deuterium
The
ions
leave
the systems,
ion source
some
Since
easily extracted. In these
a thinwith
carbon
foilvelocity.
is used that
will the
stripvacuum
away the two orbit
chamber
in aa consequence,
magnetic field,
ions suddenly
move in change
a circular
Inside
the
electrons.is As
the the
particles
fromorbit.
negative
to positive
charge
since efficiency
vacuum
electrodes,
Dees
and are chamber,
effectivelythere
bent are
out two
of the
magnetic historically
field with ancalled
almostthe
100%
extraction
andfirst
with
littlehave
activation.
the
ones
the shape of the letter D. These electrodes are hollow, which
Thethe
extracted
canfreely
eitherinside
be transported
further in
a beam
transport system or
enables
ions tobeam
move
the electrodes.
There
is aoptical
gap between
will hit a production target directly. The target is usually separated from the vacuum by
metallic foils that are strong enough to withstand the pressure difference and the heat from the
135
beam energy, as it is transferred and absorbed by the foils. The reason why two
foils are used
CHAPTER 4
the electrodes called the acceleration gap. If a voltage is applied between the
electrodes, the ions will experience the potential gradient when traversing the
gap between the electrodes. If the voltage polarity is switched at the correct rate,
the ions will be continuously accelerated when crossing the gap, thus resulting
in an increase in the ions energy and velocity. As their velocity increases, the
ions will move into a circular orbit of increasing radius. The time taken for the
ions to return to the gap is independent of their radius in accelerators <30MeV.
For the cyclotron to operate correctly, it is necessary for the frequency of the
electric field across the Dees to be the same as the frequency of the circulating
ions, so that the polarity changes upon each traversal of the ions across the Dees.
In commercial accelerators, with high beam currents of several
milliamperes, it is usual to have an internal target for radionuclide production
located inside the chamber. In accelerators with lower beam currents <100 A,
such as those dedicated for PET hospital facilities, it is more common to extract
the beam onto an external target system. The modes of extraction depend upon
whether positive or negative ions are accelerated. Extraction of positive ions is
made by using a deflector that applies a static electric field which acts upon the
particles when in the outer orbits. Some beam current is invariably lost in the
process and the deflector often becomes quite radioactive.
Modern proton/deuterium accelerators usually accelerate negative ions
that are more easily extracted. In these systems, a thin carbon foil is used
that will strip away the two orbit electrons. As a consequence, the particles
suddenly change from negative to positive charge and are effectively bent out
of the magnetic field with an almost 100% extraction efficiency and with little
activation.
The extracted beam can either be transported further in a beam optical
transport system or will hit a production target directly. The target is usually
separated from the vacuum by metallic foils that are strong enough to withstand
the pressure difference and the heat from the beam energy, as it is transferred
and absorbed by the foils. The reason why two foils are used is that the heat
produced by the beam passage has to be removed, which is facilitated by a flow
of helium gas between the foils. Helium is preferred as the cooling medium
since no induced activity will be produced in this gas.
4.3.2. Commercial production (low and high energy)
If the proton energy is >30MeV, the particles tend to be relativistic,
i.e. their mass and their cycle time in orbit increase. A constant frequency of the
accelerating electric field would cause the ions to come out of phase. This can
be compensated for either by increasing the magnetic field as a function of the
cyclotron radius (isochronic cyclotrons) or by decreasing the radiofrequency
136
during acceleration (synchrocyclotrons). Such accelerators tend to be more

complex and expensive and, for this reason, 30MeV is a typical energy for
commercial accelerators that need to have large beam currents and to be both
reliable and cost effective.
Commercial accelerators usually run beam currents of several milliamperes.
Since it is technically difficult to extract such high beam currents due to heating
problems in the separating foils, most commercial accelerators use internal
targets, i.e. targets that are placed inside the cyclotron vacuum as shown
schematically in Fig.4.13.
FIG.4.13. Schematic image of an internal target. The target material is usually thin (a few
tenths of a micrometre) and evaporated on a thicker backing plate. The target ensemble is
water cooled on the back. An advantage is that the beam is spread out over a large area which
facilitates cooling.
Many patients in nuclear medicine undergo single photon emission

computed tomography (SPECT) investigations. Besides reactor produced
99m
Tc, commercial cyclotrons commonly produce 67Ga, 111In, 123I and 201Tl. In
addition, some PET radionuclides, such as 124I, are becoming commercially
available. Increasing demand for the 68Ge/68Ga generator has also led to
commercial production of the cyclotron produced mother nuclide 68Ge. Only a
few radionuclides of medical interest require production energies above 30MeV.
A limited number of high energy accelerators with high beam currents, usually
at national physics laboratories, have the capacity for the production of, for
example, 52Fe and 61Cu and other isotopes used for research activities.
4.3.3. In-house low energy production (PET)
Commercial accelerators dedicated to PET radioisotope production
are limited both in energy (<20 MeV) and in beam current (<100 A). Many
production routes utilize gases or water as target materials and, therefore, external
137
CHAPTER 4
targets are to be preferred. Owing to the relatively low beam current, extraction is
not a problem. Since internal targets need to be taken in and out of the cyclotron
vacuum, they are not usually implemented in PET cyclotrons.
The importance of choosing the right reaction and target material is crucial
and is illustrated by the production of 18F. There are several nuclear reactions that
can be applied (Table4.3).
TABLE4.3. DIFFERENT NUCLEAR REACTIONS FOR THE PRODUCTION
OF 18F
20
Ne(d, )18F
The nascent 16F will be highly reactive. In the noble gas Ne, it will diffuse
and stick to the target walls; difficult to extract
21
Same as above; in addition, the abundance of 21Ne is low (0.27%) and needs
enrichment
19
The product and target are the same element; poor specific radioactivity
16
Cheap target but accelerators that can accelerate particles to 35MeV are
expensive and not common
Ne(p, )18F
F(p, d)18F
O(, d)18F
16
O(d, )18F
18
18
O(p, n) F
Small cross-section and no practical yields can be obtained

Expensive enriched target material but the proton energy is low (low cost
accelerator), which makes this the nuclear reaction of choice
Not only the nuclear reaction is important, but also the chemical
composition of the target. To irradiate 18O as a gas would be the purest target
(only target nuclide present) but handling a highly enriched gas in addition to
the hot-atom chemistry is complicated. Still, for some applications, this might
be the best choice. To irradiate 18O as an oxide and a solid target is possible
but the process following irradiation to dissolve the target and to chemically
separate 18F is complex, has a low yield and other elements in the oxide could
potentially contribute unwanted radioactivity. Enriched 18O water is a target of
choice as 18O is the dominant nucleus and hydrogen does not contribute to any
unwanted radioactivity. There is usually no need for target separation as water
containing 18F can often be directly used in the labelling chemistry. The target
water can also, after being diluted with saline, be injected directly into patients,
e.g. 18F-fluoride for PET bone scans. Water targets will produce 18F-fluoride for
use in stereospecific nucleophilic substitutions. An alternative production route is
neon gas production, 20Ne(d, )18F. Adding 19F2 gas to the neon as a carrier yields
18 19
F F that can be used for electrophilic substitution. Adding carrier lowers the
specific radioactivity of the labelled product.
138
A problem is the heat generated when the beam is stopped in a few

millilitres of target water. High pressure targets that force the water to remain in
the liquid phase can overcome some of these problems but production is usually
limited to beam currents <40 A. Gas and solid targets are advantageous as they
can withstand higher beam currents.
There are also several options for the production of 11C. These include:
10
B(d,n)11C, 11B(p,n)11C and 14N(p,)11C. The reactions on boron are made as
solid target irradiations while the reaction on nitrogen is a gas target application.
The routine production routes of common positron emitters associated with
PET are summarized in Table4.4.
TABLE4.4. COMMONLY USED RADIONUCLIDES IN PET
Radionuclide
15
13
11
18
Nuclear reaction
14
16
15
Yield (GBq)
N(d, n) O gas target
15
O(p, )13N liquid target
14
11
N(p, ) C gas target
18
18
O(p, n) F liquid target
40
100
Oxygen-15 is produced by deuteron bombardment of natural nitrogen

through the 14N(d,n)15O nuclear reaction. An alternative is the 15N(p, n)15O
reaction if a deuterium beam is not available. In this case, the target needs to
be enriched. In the nitrogen target, 15O-labelled molecular oxygen is produced
directly. Direct production of 11C-labelled carbon dioxide is possible by mixing
the target gas with 5% natural carbon dioxide as a carrier. Water labelled with 15O
is preferably made by processing 15O-labelled molecular oxygen.
Carbon-11 is produced by proton bombardment of natural nitrogen. By
adding a small amount of oxygen to the target gas (<0.5%), carbon dioxide
(11CO2) will be produced. Adding 5% hydrogen to the target will produce
methane (11CH4).
Liquid targets are today by far the most popular and widely used for the
production of 13N. The reaction of protons on natural water produces nitrate
and nitrite ions, which can be converted to ammonia by reduction. Water targets
can also be used to form ammonia directly with the addition of a reducing
agent, e.g. ethanol or hydrogen.
139
CHAPTER 4
4.3.4. Targetry, optimizing the production regarding

yield and impurities, yield calculations
When the nucleus is hit by an energetic particle, a complex interplay
between physical and statistical laws determines the result. Important parameters
are the entrance particle energy, the target thickness and the reaction channel
cross-sections for the particle energies in the target. Computer codes such as
ALICE and TALYS are available to calculate the size and the energy dependence
of the cross-section for a certain reaction channel but they are not easy to apply;
hence, caution should be exercised when interpreting the results from such codes.
However, a rough estimation of the irradiating particle energy can be obtained
using a well known rule of thumb in radionuclide production (illustrated in
Fig.4.14).
Cross-section (b)
As(p, n)75Se
75
As(p, 3n)73Se
75
As(p, 4n)72Se
75
Proton energy Ep (MeV)

FIG.4.14. Excitation functions of 75As(p, xn)72,73,75Se reactions. The optimal energy for the
production of 73Se is to use a proton energy of 40MeV that is degraded to 30MeV in the target.
The maximum cross-sections are found at about 10, 30 and 40MeV for the
(p, n), (p, 3n) and (p, 4n) reactions, respectively. Thus, it takes about 10MeV to
expel a nucleon, i.e. a proton of 50MeV can cover radionuclide productions that
involve the emission of about five nucleons. At low energy, there is a disturbing
production of 75Se and if excessively high proton energy is used, another
140
unwanted radionuclide impurity is produced, namely 72Se. The latter impurity

can be avoided completely by restricting the proton energy to an energy lower
than the threshold for the (p, 4n) reaction. The impurity that results from the (p, n)
reaction cannot be avoided but can beminimized by using a target thickness that
avoids the lower proton energies (having the highest (p, n) cross-sections).
Figure4.14 highlights the fact that the chosen production parameters are a
compromise. A proton range of 4030MeV uses the (p, 3n) cross-section well.
Some 72Se contamination is acceptable in order to increase the yield of 73Se. An
important factor is the half-life of 75Se (T1/2=120 d), 73Se (T1/2=7.1 h) and 72Se
(T1/2=8.5 d). Sometimes, it is possible to wait for the decay of the radioactive
contaminants. Although not the case here, sometimes a long half-life contaminant
is not a serious disadvantage. If the product half-life is long, then there may
be little product decay over the target irradiation time compared to short lived
radionuclides.
The practical set-up when undertaking radionuclide production is as follows.
A suitable As target is made and irradiated with 40MeV protons. The thickness
of the target is such that it decreases the proton energy to 30MeV. This then
gives a radioactivity yield of the desired radionuclide at the end of bombardment,
which is mainly dependent upon the beam current and the irradiation time. The
yield is usually expressed in gigabecquerels per microampere hours (GBq/Ah),
i.e. the produced radioactivity per time integrated beam current. If possible, it is
endeavoured to keep the radioactivity of the contaminants at low levels (<1%).
However, from the end of bombardment, the ratio of the product relative to any
long lived radio-contaminants begins to decrease.
4.4. RADIONUCLIDE GENERATORS
Whenever a radionuclide (parent) decays to another radioactive nuclide
(daughter), this is called a radionuclide generator. Most natural radioactivity is
produced in generator systems starting with uranium isotopes and 232Th, and
involves about fifty radioactive daughters. Several radionuclides used in nuclear
medicine are produced by generator systems such as the 99Mo production of
99m
Tc, which subsequently decays to 99Tc. The extremely long half-life of 99Tc
(T1/2=2.1105 a) means that 99mTc can be safely used as a clinical isotope
without any radiological concerns. In other nuclides, the creation of a radioactive
nuclide may be more important, e.g. the positron emitter 52Fe (T1/2=8 h) decays to
52
Mn (T1/2=21min) which is also a positron emitter. Furthermore, radionuclides
used in therapy may themselves be generators such as 211At (T1/2=7 h) decaying
to 211Po (T1/2=0.5 s) or 223Ra, which generates a series of relatively short lived
radioactive daughters in situ.
141
CHAPTER 4
When talking about generators in nuclear medicine, a special case is usually

considered in which a long lived mother generates a short lived daughter, which
after labelling is administrated to the patient. Generally, this is a practical way
to deliver short lived radionuclides to hospitals which otherwise, for logistical
reasons, would not have been possible. The half-life should be sufficiently long
so that the radionuclide can be delivered to hospitals, and provide the radioactive
product for a number of patients over days or weeks. A typical example is the
99
Mo/99mTc generator (Fig.4.15), which produces the most used radionuclide in
nuclear medicine. The half-life of the parent (2.7 d) is adequate for transport and
delivery, and the daughter has a suitable half-life (6 h) for patient investigations.
The generator is used for about two to three half-lives of the parent (1 week) after
which time it is renewed.
FIG.4.15. Elution of a 99Mo/99mTc generator. The generator has a nominal activity of

1000MBq on day 0 (Monday). It is eluted daily, five times a week, yielding 1000, 780, 600,
470 and 360MBq.
4.4.1. Principles of generators

Generator systems require that the parent is a reactor or accelerator
produced by the methods described above and that the daughter radionuclide of
interest can easily be separated from the parent. The 99Mo/99mTc generator exhibits
these characteristics. Most commercial generators use column chromatography,
in which 99Mo is adsorbed onto alumina. Eluting the immobilized 99Mo on the
column with physiological saline elutes the soluble 99mTc in a few millilitres of
142
liquid. In fact, most generators in nuclear medicine use ion exchange columns in
much the same way due to its simplicity of handling.
In generator systems, the daughter radionuclide is formed at the rate at
which the parent decays, PNP. It also decays at the same rate, DND, as the parent,
once a state of transient equilibrium has been reached. The equations that describe the
relationship between parent and daughter are provided in Chapter1.
Another generator of increasing importance is 68Ge, which has a half-life
of 271 d that produces a short lived positron emitter 68Ga (T1/2=68min). This is
produced as a +3 ion that can be tagged, using a chelating agent such as DOTA,
to small peptides, e.g. 68Ga-DOTATOC. Owing to the long half-life of the mother,
the generator can be operated for up to two years and can be eluted every 5 h. One
problem with such a long lived generator is keeping it sterile, and furthermore,
the ion exchange material is exposed to high radiation doses that may reduce the
elution efficiency and the quality of the product.
The 90Sr/90Y generator is used to produce the therapeutic radionuclide 90Y.
This generator is not distributed to hospitals but is operated in special laboratories
on account of radiation protection considerations associated with the long lived
parent. The daughter, 90Y, has a half-life of 2.3 d which is adequate for transport
of the eluted 90Y to distant hospitals.
81
Rb (4.5 h)/81mKr (13.5 s) for ventilation studies and 82Sr (25.5 d)/82Rb (75 s)
for cardiac PET studies are examples of other generators with special requirements
due to the extremely short half-life of the eluted product. Recently,
generator systems producing emitters for therapy have become available,
e.g. 225Ac (10 d)/213Bi (45.6min).
4.5. RADIOCHEMISTRY OF IRRADIATED TARGETS
During target irradiation, a few atoms of the wanted radionuclide are
produced within the bulk target material. The energy released in a nuclear
reaction is large relative to the electron binding energies and the radionuclide is,
therefore, usually born almost naked with no or few orbit electrons. This hot
atom will undergo chemical reactions depending on the target composition. In
a gas or liquid target, these hot atom reactions may even cause the activity to
be lost in covalent bonds to the target holder material. During irradiation, the
target is also heated and its structure and composition may change. A pressed
powder target may be sintered and become more ceramic, which makes it more
difficult to dissolve. The target may melt and the radioactivity may diffuse in the
target and even possibly evaporate. In designing a separation method, all of these
factors have to be considered. Fast, efficient and safe methods are required to
143
CHAPTER 4
separate the few picograms of radioactive product from the bulk target material
which is present in gram quantities.
Separation of the radionuclide already starts in the target as demonstrated
in the production of 11CO2. Carbon-11 is produced in a (p, ) reaction on
nitrogen gas. To enable the production of CO2, some trace amounts of oxygen
gas (0.10.5%) are added. However, at low beam currents, mainly CO will be
formed, since the target will not be heated. At high beam currents, the CO will
be oxidized to the chemical form CO2. The separation, made by letting the target
through a liquid nitrogen trap, is simple and efficient. By adding hydrogen gas
instead, the product will be CH4.
The skill in hot-atom chemistry is to obtain a suitable chemical form of
the radioactive product, especially when working with gas and liquid targets.
Solid targets are usually dissolved and chemically processed to obtain the wanted
chemical form for separation.
4.5.1. Carrier-free, carrier-added systems
The concept of specific activity a, i.e. the activity per mass of a preparation,
is essential in radiopharmacy. If 100% of the product contains radioactive atoms,
often called the theoretical a, then the relationship between the activity A in
becquerels and the number of radioactive atoms N is given by N=A/, where is
the decay constant (1/s). The decay constant can be calculated from the half-life
T1/2 in seconds as =ln(2)/T1/2.
The specific activity a expressed as activity per number of radioactive atoms
is then A/N==ln(2)/T1/2. For a short lived radionuclide, a will be relatively
large compared to a long lived isotope. For example, a for 11C (T1/2=20min) is
1.5108 times larger than for 14C (T1/2=5730 a).
The specific activity a expressed in this way is a theoretical value that is
rarely obtained in practical work. When producing 11C, the target gas and target
holder will contain stable carbon that will dilute the radioactive carbon as well
as compete in the labelling process afterwards. A more empirical way to define
a is to divide the activity by the total mass of the element under consideration.
This value for 11C will usually be a few thousand times lower than the theoretical
value, while the production of 14C can come closer to the theoretical a.
In the labelling process, a is usually expressed as the activity per number
of molecules (a sum of labelled and unlabelled molecules). Instead of using the
number of atoms or molecules, it is common to use the mole concept by dividing
N by Avogadros number (NA=6.0221023). A common unit for a is then
gigabecquerels per micromole.
If the radioactive atoms are produced and separated from the target without
any stable isotopes, the process is said to be carrier-free. If stable isotopes are
144
introduced as being a contaminant in the target or in the separation procedure,

the process is said to have no carrier added, i.e. no stable isotope is deliberately
added. Both of these processes usually give a high final a. However, it may be
necessary to use a target of the same element or it may be necessary to add extra
mass of the same element in order for the separation process to work. In this case,
carrier is added deliberately and a will usually be low.
It should be noted that a carrier does not necessarily need to be of the same
element. When labelling a radiopharmaceutical with a chelator and metal ions,
any ion fitting into the chelator will compete. An example is labelling a peptide
with 111In, when the activity will usually be delivered as InCl3 in a weak acid. By
sampling the activity with a stainless steel needle, Fe ions will be released and
will probably completely ruin the labelling process by outnumbering the 111In
atoms.
4.5.2. Separation methods, solvent extraction,
ion exchange, thermal diffusion
After irradiation, the small amount of desired radioactivity (of the order of
nanomoles) usually needs to be separated from the bulk of the target in a suitable
form for the following labelling process and at high a. The separation time should
be related to the half-life of the radionuclide and should take at most one half-life.
Solid targets usually have to be dissolved, which is simple for salts such as NaI
but more complicated for, for example, Ni foils where boiling aqua regia may
have to be used. To speed up this process, the Ni foil can be replaced by a pressed
target of Ni powder that will increase the metal surface and will speed up the
dissolving process.
In general, two principles are used: liquid extraction and ion exchange. In
liquid extraction, usually two liquids that do not mix are used, e.g. water and an
organic solvent. The target element and the produced activity of another element
should have different relative solubility in the liquids. The two liquids and the
dissolved target are mixed by shaking, after which two phases are formed. The
phase with a high concentration of the wanted radioactive product is sampled
and is usually separated again one or more times to reduce the target mass in that
fraction. The relative solubility can be optimized by varying the pH or by adding
a complexing agent.
In the ion exchange mechanism, an ion in the liquid phase (usually an
aqueous phase) is transferred to a solid phase (organic or ceramic material). To
maintain the charge balance, a counter ion is released from the solid phase. This
ion may be a hydrogen ion. In the ion exchange mechanism, the distribution ratio
is often a function of the pH. Furthermore, complexing agents can be used to
modify the distribution ratio. The dissolved target is adjusted to obtain the right
145
CHAPTER 4
pH and other separation conditions, and is then put on to a column containing the
ion exchange material. The optimal separation conditions would be that the small
mass of desired radioactivity but not the bulk target material sticks to the column.
The column can then be small, and after washing and change of pH, the desired
activity can be eluted in a small volume. Under other conditions, large amounts
of ion exchange material have to be used to prevent saturation of binding sites
and leakage of the target material. This also means that large liquid volumes have
to be used, implying poorer separation. The two techniques are often performed
together by using liquid extraction to reduce the target mass, after which ion
exchange is used to make the final separation.
Occasionally, thermal separation techniques may be applied, which have
the advantage that they do not destroy the target (important when expensive
enriched targets are used) and that they lend themselves to automation. As an
example of such dry methods, the thermal separation of 76Br (T1/2=16 h) is
described. The target is Cu276Se, a selenium compound that can withstand some
heat. The nuclear reaction used is 76Se(p, n)76Br.
The process is as follows:
(a) The target is placed in a tube and heated, under a stream of argon gas, to
evaporate the 76Br activity by dry distillation (Fig.4.16);
(b) A temperature gradient is applied to separate the deposition areas of 76Br and
traces of co-evaporated selenide in the tube by thermal chromatography;
(c) The 76Br activity deposited on the tube wall is dissolved in small amounts
of buffer or water.
Argon
FIG.4.16. A schematic description of the 76Br separation equipment: (1) furnace, (2) auxiliary
furnace, (3) irradiated target, (4) deposition area of selenium, (5) deposition area of 76Br,
(6) gas trap.
146
Separation yields of 6070% are achieved by this method, with a separation

time of about 1 h. Since dry distillation permits the extraction of radiobromine
without destroying the target, the Cu2Se targets are reusable. Considering the
rather expensive 76Se-enriched target material, this is a practical prerequisite for
this type of production. The chemical form of the 76Br activity after separation,
analysed by ion exchange high performance liquid chromatography and thin-layer
chromatography, was almost exclusively found to be bromide.
4.5.3. Radiation protection considerations and hot-box facilities
Besides the desired activity, the irradiated target usually contains a number
of other radionuclides of varying elements, half-lives and energies. The
presence of such contaminants needs to be taken into account when planning
radiopharmaceutical labelling. An example is the production of 35S using the
reaction 35Cl(n, p)35S. At first glance, NaCl would be a suitable target due to the
low atomic weight of sodium, a single isotope (23Na) and a salt that is easy to
dissolve. However, 23Na has a huge thermal neutron cross-section for producing
24
Na, which has a half-life of 15h and abundant energies up to 2.75MeV. This
target would be extremely hot, demanding lead protection more than 30cm thick.
If instead KCl were used, the emitted radiation energy would be substantially
lower and decay times shorter.
After irradiation, the target is usually stored before processing to allow
any short lived radionuclides to decay. Depending on the half-life, this cooling
period can be fromminutes to months, but should not exceed one half-life of the
desired radionuclide. The place used for this depends on the source activity and
the energy and abundance of the emissions. Separation of fairly pure and
emitters may require just some distance and some plastic shielding, and can be
performed in a standard fume hood, while targets with a high emission need
significant lead shielding.
Handling reactor or accelerator produced radioactivity of the order of
several hundred gigabecquerels requires adequate radiation protection, usually
in the form of lead shields, hot-boxes, lead shielded fume hoods and laminar air
flow benches. Typical lead thicknesses required by common radionuclides are
indicated in Table4.5.
The radioactive target and the radionuclide separation is often the first
step in labelling a radiopharmaceutical. The hot-box then has to fulfil the
requirements both to protect the operator from the radiation and to protect the
pharmaceutical from the surroundings. The first step usually requires a negative
pressure hood to prevent eventual airborne radioactivity to leak out into the
laboratory, while the second step requires a high positive pressure to be applied
across the pharmaceutical to avoid contact with less pure air from the laboratory.
147
CHAPTER 4
TABLE4.5. DOSE RATES AND LEAD SHIELDING REQUIRED FOR

DIFFERENT RADIONUCLIDES, DETERMINED BY THE GAMMA
RADIATION ABUNDANCE AND ENERGYa
Dose rate (mSv/h) at 1 m per TBq
99m
Tc
18
111
In
81
18
124
135
117
Thickness of lead shield (cm) giving 1 Sv/h

TBq
99m
Tc
111
In
18
124
0.1
0.28
1.0
5.8
20
1.0
0.36
1.3
7.1
22
10.0
0.43
1.6
8.5
27
Calculations made with RadProCalculator (http://www.radprocalculator.com/).
These contradictory conditions are usually handled by having a box in the box,
i.e. the pharmaceutical is processed in a closed facility at over pressure placed in
the hot-box having low pressure. The classical hot-box design, with manipulators
to manually process the radioactivity remotely, as seen in Fig.4.17, is gradually
being replaced by lead protected chambers housing an automatic chemistry
system or a chemical robot making the pharmaceutical computer controlled.
FIG. 4.17. Examples of modern hot-box designs (courtesy of Von Gahlen Nederland B.V.).
148
FIG. 4.17. Examples of modern hot-box designs.
CHAPTER 5
STATISTICS FOR RADIATION MEASUREMENT
M.G. LTTER
University of the Free State,
Bloemfontein, South Africa
5.1. SOURCES OF ERROR IN NUCLEAR MEDICINE MEASUREMENT
Measurement errors are of three general types: (i) blunders, (ii) systematic
errors or accuracy of measurements, and (iii) random errors or precision of
measurements.
Blunders produce grossly inaccurate results and experienced observers
easily detect their occurrence. Examples in radiation counting or measurements
include the incorrect setting of the energy window, counting heavily contaminated
samples, using contaminated detectors for imaging or counting, obtaining
measurements of high activities, resulting in count rates that lead to excessive
dead time effects, and selecting the wrong patient orientation during imaging.
Although some blunders can be detected as outliers or by duplicate samples and
measurements, blunders should be avoided by careful, meticulous and dedicated
work. This is especially important where results will determine the diagnosis or
treatment of patients.
Systematic errors produce results that differ consistently from the correct
results by some fixed amount. The same result may be obtained in repeated
measurements, but overestimating or underestimating the true value. Systematic
errors are said to influence the accuracy of measurements. Measurement results
having systematic errors will be inaccurate or biased. Examples of a systematic
error are:
When an incorrectly calibrated ionization chamber is used for measurement
of radiation dose.
When during thyroid uptake studies with 123I the count rate of the reference
standard results in dead time losses. The percentage of thyroid uptake will
be overestimated.
When in sample counting the geometry of samples and the position within
the detector are not the same as in the reference sample.
149
CHAPTER 5
When during blood volume measurements the tracer leaks out of the blood
compartment. The theory of the method assumes that the tracer will stay
in the blood compartment. The leaking of the tracer will consistently
overestimate the measured blood volume.
When in calculation of the ventricular ejection fraction during gated
blood pool studies the selected background counts underestimate the true
ventricular background counts, the ejection fraction will be consistently
underestimated.
Measurement results affected by systematic errors are not always easy to
detect, since the measurements may not be too different from the expected results.
Systematic errors can be detected by using reference standards. For example,
radionuclide standards calibrated at a reference laboratory should be used to
calibrate source calibrators to determine correction factors for each radionuclide
used for patient treatment and diagnosis.
Measurement results affected by systematic errors can differ from the true
value by a constant value and/or by a fraction. Using golden standard reference
values, a regression curve can be calculated. The regression curve can be used to
convert systematic errors to a more accurate value. For example, if the ejection
fraction is determined by a radionuclide gated study, it can be correlated with the
golden standard values.
Random errors are variations in results from one measurement to the next,
arising from actual random variation of the measured quantity itself, as well as
physical limitations of the measurement system.
Random error affects the reproducibility, precision or uncertainty in the
measurement. Random errors are always present when radiation measurements
are performed because the measured quantity, namely the radionuclide decay,
is a random varying quantity. The random error during radiation measurements
introduced by the measured quantity, that is the radionuclide decay, is
demonstrated in Fig.5.1. Figure5.1 shows the energy spectrum of a 57Co source
in a scattering medium and measured with a scintillation detector probe. The
energy spectrum represented by square markers is the measured energy spectrum
with random noise due to radionuclide decay. The solid line spectrum represents
the energy spectrum without random noise. The variation around the solid line of
the data points, represented by markers, is a result of random error introduced by
radionuclide decay.
The influence of the random error of the measurement system introduced
by the scintillation detector is also demonstrated in Fig.5.1. Cobalt-57 emits
photons of 122keV and with a perfect detection system all of the counts are
expected at 122keV. The measurements are, however, spread around 122keV
as a result of the random error introduced by the scintillation detector during the
150
detection of each photon. When a photon is detected with the scintillation

detector, the number of charge carriers generated will vary randomly. The varying
number of charge carriers will cause varying pulse heights at the output of the
detector and this variation determines the spread around the true photon energy
of 122keV. The width of the photopeak determines the energy resolution of the
detection system.
FIG.5.1. Energy spectrum of a

scintillation detector.
57
Co source in a scattering medium obtained with a
Random errors also play a significant role in radionuclide imaging. Here,

the random error as a result of the measured quantity, namely radionuclide decay,
will significantly influence the visual quality of the image. This is because the
number of counts acquired in each pixel is subject to random error. It is shown that
the relative random error decreases as the number of counts per pixel increases.
The visual effect of the random error as a result of the measured quantity is
demonstrated in Fig.5.2. Technetium-99m planar bone scans (acquired on a
256256 matrix) were acquired with a scintillation camera. Image acquisition
was terminated at a total count of 21, 87 and 748 kcounts. When the total number
of counts per image are increased, the counts per pixel increase and the random
error decreases, resulting in improved visual image quality. As the accumulated
counts are increased, the ability to visualize anatomical structures and, more
importantly, tumour volumes, significantly increases. The random error
introduced by the measuring system or imaging device, such as a scintillation
camera, also influences image quality. This is as a result of the energy resolution
and intrinsic spatial resolution of imaging devices that are influenced by random
errors during the detection of each photon. The energy resolution of the system
151
CHAPTER 5
will determine the ability of the system to reject lower energy scattered photons
and improve image contrast.
FIG.5.2. The influence of random error as a result of radionuclide decay or counting statistics
is demonstrated for imaging. Technetium-99m posterior planar bone images (256256) using
a scintillation camera were acquired to total counts of 21, 87 and 748 kcounts.
It is possible for a measurement to be precise (small random error) but

inaccurate (large systematic error), or vice versa. For example, for the calculation
of the ejection fraction during gated cardiac studies, the selection of the
background region of interest (ROI) will be exactly reproducible when a software
algorithm is used. However, if the algorithm is such that the selected ROI does
not reflect the true ventricular background, the measurement will be precise but
inaccurate. Conversely, individual radiation counts of a radioactive sample may
be imprecise because of the random error, but the average value of a number of
measurements will be accurate, representing the true counts acquired.
Random errors are always present and play a significant role in radiation
counting and imaging. It is, therefore, important to analyse the random errors to
determine the associated uncertainty. This is done using methods of statistical
analysis. The remainder of the chapter describes methods of analysis.
The analysis of radiation measurements and imaging forms a subgroup
of general statistical analysis. In this chapter, the focus is on statistical analysis
for radiation counting and imaging measurements, although some methods
described will be applicable to a wider class of experimental data as described in
Sections 5.2, 5.3 and 5.5.
152
5.2. CHARACTERIZATION OF DATA

5.2.1. Measures of central tendency and variability
5.2.1.1. Dataset as a list
Two measurements of the central tendency of a set of measurements are
the mean (average) and median. It is assumed that there is a list of N independent
measurements of the same physical quantity:
x1, x2, x3, .xixN
It is supposed that the dataset is obtained from a long lived radioactive
sample counted repeatedly under the same conditions with a properly operating
counting system. As the disintegration rate of the radioactive sample undergoes
random variations from one moment to the next, the number of counts recorded
in successive measurements is not the same as the result of random errors in the
measurement.
The experimental mean x e of the set of measurements is defined as:
xe =
x 1 + x 2 + ... + x N
(5.1)
N
N
x
i =1
(5.2)
The following procedure is followed to obtain the median. The list of

measurements must first be sorted by size. The median is the middlemost
measurement if the number of measurements is odd and is the average of the
two middlemost measurements if the number of measurements is even. For
example, to obtain the median of five measurements, 7, 13, 6, 10 and 14, they are
first sorted by size: 6, 7, 10, 13 and 14. The median is 10. The advantage of the
median over the mean is that the median is less affected by outliers. An outlier is
a blunder and is much greater or much less than the others.
The measures of variability, random error and precision of a list of
measurements are the variance, standard deviation and fractional standard
deviation, respectively.
153
CHAPTER 5
The variance e2 is determined from a set of measurements by subtracting

the mean from each measurement, squaring the difference, summing the squares
and dividing by one less than the number of measurements:
e2 =
( x 1 x e ) 2 + ( x 2 x e ) 2 + ........ + ( x N x e ) 2
N 1
N
= N11
(x x )
i
(5.3)
i =1
where N is the total number of measurements and x e is the experimental mean.

The standard deviation e is the square root of the variance:
e = e 2 (5.4)
The fractional standard deviation eF (fractional error or coefficient of

variation) is the standard deviation divided by the mean:
eF =
e
(5.5)
x e (5.5)
The fractional standard deviation is an important measure to evaluate

variability in measurements of radioactivity. The inverse of the fractional
standard deviation 1/eF in imaging is referred to as the signal to noise ratio.
5.2.1.2. Dataset as a relative frequency distribution function
It is often convenient to represent the dataset by a relative frequency
distribution function F(x). The value of F(x) is the relative frequency with which
the number appears in the collection of data in each bin. By definition:
F ( x) =
number of occurrences of the value x in each bin

number of measurements (N )
(5.6)
The distribution is normalized, that is:
(5.7)
F (x) = 1 (5.7)
x =0
154
As long as the specific sequence of numbers is not important, the complete

data distribution function represents all of the information in the original dataset
in list format.
Figure5.3 illustrates a demonstration of the application of the relative
frequency distribution. The scintillation counter measurements appear noisy due
to the random error as a result of the measured quantity (the radionuclide decay)
(Fig.5.3(a)). The measurements fluctuate randomly above and below the mean
of 90 counts. A histogram (red bars) of the relative frequency distribution of
the fluctuations in the measurements can be constructed by plotting the relative
frequency of the measured counts (Fig.5.3(b)). The x axis represents the range of
possible counts that were measured in each of the bins with a six count interval.
The y axis represents the relative frequencies with which the particular count
values occur. The most common value, that is 26% of the measurements, is near
the mean of 90 counts. The values of the other measurements are substantially
higher or lower than the mean. The measured frequency distribution histogram
agrees well with the expected calculated normal distribution (blue curve).
Measurement counts
(a) Measured counts
(b) Relative frequency distribution

Measured counts
Calculated frequency distribution
Measured counts
Average
Measurement number
Counts
FIG.5.3. One thousand measurements were made with a scintillation counter. (a)The graph
shows the variations observed for the first 50 measurements. (b)The graph (red bars) shows
the histogram of the relative frequency distribution for the measurements as well as the
expected calculated frequency distribution.
The relative frequency distribution is a useful tool to provide a quick visual

summary of the distribution of measurement values and can be used to identify
outliers such as blunders or the correct functioning of equipment.
155
CHAPTER 5
Three measurements of the central tendency for a frequency distribution

are the mean (average), median and mode:
The mode of a frequency distribution is defined as the most frequent value
or the value at the maximum probability of the frequency distribution.
The median of a frequency distribution is the value at which the integral of
the frequency distribution is 0.5; that is, half of the measurements will be
smaller and half will be larger than the median.
The experimental mean x e using the frequency distribution function
can be calculated. The experimental mean is obtained by calculating the first
moment of the frequency distribution function. The equation for calculating the
mean can also be derived from the equation for calculating the mean for data
in a list (Eq.(5.2)). The sum of measurements
is equal to the sum of the
i =1
measurements in each bin in the frequency distribution function. The sum of the
measurements for each bin is obtained by multiplying the value of the bin i and
the number of occurrences of the value xi.
N
xe =
i =1
N
N
[value of bin (i)] [number of occurences of the value x i ]

N
i =1
x F (x )
(5.8)
(5.8)
i =1
The experimental sample variance can be calculated using the frequency

distribution function:
e2 =
N
N 1
(x x ) F (x )
i
(5.9)
(5.9)
i =1
The standard deviation and the fractional standard deviation are given by
Eqs(5.4) and (5.5).
The frequency distribution provides information and insight on the
precision of the experimental sample mean and of a single measurement.
Figure5.3 demonstrates the distribution of counting measurements around the
156
true mean ( x t ) . The value of the true mean is not known but the experimental
sample mean ( x e ) can be used as an estimate of the true mean ( x t ) :
( x t ) ( x e ) (5.10)
(5.10)
In routine practice, it is often impractical to obtain multiple measurements

and one must be satisfied with only one measurement. This is especially the
case during radionuclide imaging and nuclear measurements on patients. The
frequency distribution of the measurements will determine the precision of a
single measurement as an estimate of the true value. The probability that a single
measurement will be close to the true mean depends on the relative width or
dispersion of the frequency distribution curve. This is expressed by the variance
2 (Eq.(5.9)) or standard deviation of the distribution. The standard deviation
is a number such that 68.3% of the measurement results fall within of the true
mean x t .
Given the result of a given measurement x, it can be said that there is a
68.3% chance that the measurement is within the range x. This is called the
68.3% confidence interval for the true mean x t . There is 68.3% confidence that
x t is in the range x. Other confidence intervals can be defined in terms of the
standard deviation . They are summarized in Table5.1. The 50% confidence
interval (0.675) is referred to as the probable error of the true mean x t .
TABLE5.1. Confidence Levels in radiation Measurements
Range
Confidence level for true mean x t
x t 0.675
50.0
x t 1.000
68.3
x t 1.640
90.0
x t 2.000
95.0
x t 3.000
99.7
5.3. STATISTICAL MODELS

Under certain conditions, the distribution function that will describe
the results of many repetitions of a given measurement can be predicted. A
measurement is defined as counting the number of successes x resulting from a
given number of trials n. Each trial is assumed to be a binary process in that only
157
CHAPTER 5
two results are possible: the trial is either a success or not. It is further assumed
that the probability of success p is constant for all trials.
To show how these conditions apply in real situations, Table5.2 gives
four separate examples. The third example gives the basis for counting nuclear
radiation events. In this case, a trial consists of observing a given radioactive
nucleus for a period of time t. The number of trials n is equivalent to the number
of nuclei in the sample under observation, and the measurement consists of
counting those nuclei that undergo decay. We identify the probability of success
as p. For radioactive decay:
p = (1 e t ) (5.11)
where is the decay constant of the radionuclide.

The fifth example demonstrates the uncertainty associated with the energy
determination during scintillation counting. The light photons generated in the
scintillator following interaction with an incoming ray will eject electrons at the
photocathode of the photomultiplier tube (PMT). Typically, one electron ejected
for every five light photons results in a probability of success of 1/5.
Table5.2. Examples of Binary Processes
Trial
Definition of success
Probability of success p
Tossing a coin
Heads
1/2
Rolling a die
A six
1/6
Observing a given radionuclide

for time t
The nucleus decays during

observation
(1 e lt )
Observing a given ray over

a distance x in an attenuating
medium
The ray interacts with the

medium during observation
(1 e x )
Observing light photons

generated in a scintillator
An electron is ejected from the

photocathode
1/5
5.3.1. Conditions when binomial, Poisson and

normal distributions are applicable
Three statistical models are used: the binomial distribution, the Poisson
distribution and the Gaussian or normal distribution. Figure5.4 shows the
158
distribution for the three models. The distributions were generated by using a
Microsoft Office Excel spreadsheet.
5.3.1.1. Binomial distribution
This is the most general model and is widely applicable to all constant
p processes (Fig.5.4). Binomial distribution is rarely used in nuclear decay
applications. One example in which the binomial distribution must be used is
when a radionuclide with a very short half-life is counted with a high counting
efficiency.
x
P(x)
P(x)
x
p
P(x)
P(x)
FIG.5.4. Probability distribution models for successful event probability p=0.4 and
p=0.0001 for x = 10 and x =100, respectively.
5.3.1.2. Poisson distribution

The model is a direct mathematical simplification of the binomial
distribution under conditions that the event probability of success p is small
(Fig.5.4). For nuclear counting, this condition implies that the chosen observation
time is small compared to the half-life of the source, or that the detection
159
CHAPTER 5
efficiency is low. The Poisson distribution is an important distribution. When

the success rate is low, the true experimental distribution is asymmetric and a
Poisson distribution must then be used since the normal distribution is always
symmetrical. This is demonstrated in Fig.5.4 for p=0.0001 and x =10.
5.3.1.3. Gaussian or normal distribution
The third important distribution is the normal or Gaussian, which is
a further simplification if the mean number of successes x is relatively large
(>30). At this level of success, the experimental distribution will be symmetrical
and can be represented by the Gaussian distribution (Fig.5.4). The Gaussian
model is widely applicable to many applications in counting statistics.
It should be emphasized that the distribution of all of the above models
becomes identical for processes with a small individual success probability p
and with a large enough number of trials such that the expected mean number of
successes x is large. This is demonstrated in Fig.5.4 for p=0.0001 and x =100.
5.3.2. Binomial distribution
Binomial distribution is the most general of the statistical models
discussed. If n is the number of trials for which each trial has a success probability
p, then the predicted probability of counting exactly x successes is given by:
P( x) =
n!
p x (1 p) nx (5.12)
(5.12)
(n x)! x !
P(x) is the predicted probability distribution function, and is defined only

for integer values of n and x. The value of n! (n factorial) is the product of integers
up to n, that is 123 n. The values of x! and (n x)! are similarly
calculated.
The properties of the binomial distribution are as follows:
The distribution is normalized:
n
P( x) = 1
(5.13)
(5.13)
x =0
The mean value x of the distribution using Eq.(5.8) is given by:

n
x=
xP(x)
x =0
160
(5.14)
(5.14)
If Eq.(5.12) is substituted for P(x), the mean value x of the distribution is

given by:
x = pn (5.15)
(5.15)
The sample variance for a set of experimental data has been defined by
Eq.(5.9). By analogy, the predicted variance 2 is given by:
2 =
(x x )
P( x) (5.16)
(5.16)
x =0
If Eq.(5.12) is substituted for P(x), the predicted variance 2 of the

distribution will be:
2 = np(1 p) (5.17)
(5.17)
If Eq.(5.15) is substituted for np:

2 = x (1 p) (5.18)
(5.18)
The standard deviation is the square root of the predicted variance 2:

= np(1 p) = x (1 p)
(5.19)
(5.19)
The fractional standard deviation F is given by:

F =
np(1 p)
(1 p)
=
np
np
F =
x (1 p)
=
x
(1 p)
(5.20)
(5.20)
x
Equation (5.19) predicts the amount of fluctuation inherent in a given

binomial distribution in terms of the basic parameters, namely the number of
trials n and the success probability p, where x = pn .
5.3.2.1. Application example of binomial distribution
The operation of a scintillation detector (Section 6.4) is considered. It
consists of a scintillation crystal mounted on a PMT in a light tight construction.
Firstly, when a ray interacts with the crystal, it generates n light photons.
161
CHAPTER 5
Secondly, the light photons then eject x electrons from the photomultiplier
photocathode. Thirdly, these electrons are then multiplied to form a pulse that
can be further processed. For each ray that interacts with the scintillator, the
number of light photons n, electrons ejected x and multiplication vary statistically
during the detection of the different rays. This variation determines the energy
resolution of the system.
In this example, the second stage is illustrated, that is the ejection of
electrons from the photocathode. The variation or the standard deviation and
fractional standard deviation for the number of electrons x that are ejected can
be calculated using the binomial distribution as is given by Eqs(5.19) and (5.20).
The typical values for a scintillation counter are as follows. It is assumed
that the 142keV rays emitted by 99mTc are being counted. It is further assumed
that it uses 100 eV to generate a light photon in the scintillation crystal when a
ray interacts with the crystal. Therefore, if all of the energy of a single 142keV
photon is absorbed, n=142 000/100=1420 light photons will be emitted. It is
assumed that these light photons fall on the photocathode of the PMT to generate
x electrons for each ray absorbed. It is further assumed that five light photons
are required to eject one electron.
For the binomial distribution, the probability of a light photon ejecting
an electron is p=1/5 and the number of trials n will be the number of light
photons generated for each ray. This will be 1420. Equation (5.15) can be used
to calculate the predicted mean number of electrons ejected for each ray:
x = pn =
1
1420 = 284 electrons
5
The standard deviation (Eq.(5.19)) and relative standard deviation

(Eq.(5.20)) can be calculated using the binomial distribution:
= x (1 p) = 284(1 1 / 5) = 15
and
F =
(1 p)
(1 1 / 5)
(5.20)
=
= 0.053 (5.21)
x
284
Therefore, the contribution to the overall standard deviation at the electron

ejection stage at the photocathode is 5.3%. The variation in the number of
electrons will influence the pulse height obtained for each ray. The variation
in the pulse height during the detection of rays will determine the width of the
162
photopeak (Fig.5.1) and the energy resolution of the system (Sections 5.7.1 and
6.4).
5.3.3. Poisson distribution
Many binary processes can be characterized by a low probability of
success for each individual trial. This includes nuclear counting and imaging
applications in which large numbers of radionuclides make up the sample or
number of trials, but a relatively small fraction of these give rise to recorded
counts. Similarly, during imaging, many rays are emitted by the administered
imaging radionuclide, for every one that interacts with the tissue. In addition,
during nuclear counting, many rays strike the detector for every single recorded
interaction.
Under these conditions, the approximation that the probability p is small
(p 1) will hold and some mathematical simplifications can be applied to the
binomial distribution. The binomial distribution reduces to the form:
P( x) =
( pn) x e pn
x!
(5.21)(5.22)
The relation pn = x holds for this distribution as well as for the binomial
distribution:
P( x) =
( x ) x e x
x!
(5.22)
(5.23)
Equation (5.23) is the form of the Poisson distribution.

For the calculation of binomial distribution, two parameters are required:
the number of trials n and the individual success probability p. It is noted from
Eq.(5.23) that only one parameter, the mean value x , is required. This is a very
useful simplification because using only the mean value of the distribution, all
other values of the Poisson distribution can be calculated. This is of great help
for processes in which the mean value can be measured or estimated, but for
which there is no information about either the individual probability or size of the
sample. This is the case in nuclear counting and imaging.
The properties of the Poisson distribution are as follows. The Poisson
distribution is a normalized frequency distribution (seeEqs(5.6) and (5.7)):
n
P(x) = 1 (5.24)
x =0
163
CHAPTER 5
The mean value or first moment for the Poisson distribution is calculated
by inserting the Poisson distribution (Eq.(5.22)) into the equation to calculate the
mean for a frequency distribution (Eq.(5.8)):
x=
xP(x) = pn
x =0
(5.24)
(5.25)
This is the same result as was obtained for the binomial distribution.
The predicted variance of the Poisson distribution differs from that of the
binomial distribution and can be derived from Eqs(5.9) and (5.22):
2 =
(x x )
P( x) = pn (5.26)
x =0
From the result of Eq.(5.26), the predicted variance is reduced to the

important general equation:
2 = x (5.27)
The predicted standard deviation is the square root of the predicted variance
(Eq.(5.4)):
= 2 = x (5.28)
The predicted standard deviation of any Poisson distribution is just the

square root of the mean value that characterizes the same distribution.
The predicted fractional standard deviation F (fractional error or coefficient
of variation) is the standard deviation divided by the mean (Eq.(5.5)):
F =
1
1
=
= (5.29)
x
x
The fractional standard deviation is the inverse of the square root of the
mean value of the distribution.
Equations (5.28) and (5.29) are important equations and frequently find
application in nuclear detection and imaging.
164
5.3.4. Normal distribution

The Poisson distribution holds as a mathematical simplification to the
binomial distribution within the limit p < 1. If, in addition, the mean value of the
distribution is large (>30), additional simplification can generally be carried out
which leads to a normal or Gaussian distribution:
( xx ) 2
2 x
e
P( x) =
2 x
(5.30)
The distribution function is only defined for integer values of x.

Figure5.4 for x = 100 and p=0.0001 demonstrates that for these values
the normal distribution is identical to the Poisson and binomial distributions. The
normal distribution is always symmetrical or bell-shaped (Fig.5.4). It shares
the following properties with the Poisson distribution:
It is normalized (seeSection 5.2.1.2 and Eqs(5.6) and (5.7)):
n
P(x) = 1 (5.31)
x =0
The distribution is characterized by a single parameter x = pn .

The predicted variance of the normal distribution is given by the mean of x:
2 = x (5.32)
The predicted standard deviation is the square root of the predicted variance
(Eq.(5.4)):
= 2 = x (5.33)
The predicted fractional standard deviation F (fractional error or coefficient

of variation) is the standard deviation divided by the mean (Eq.(5.5)):
F=
1
1
=
x
(5.34)
The fractional standard deviation is the inverse of the square root of the
mean value of the distribution.
165
CHAPTER 5
5.3.4.1. Continuous normal distribution: confidence intervals

In experiments where the sample size is small, there are only a few discrete
outcomes. As the sample size increase, so does the number of possible sample
outcomes. As the sample size approaches infinity, there is, in effect, a continuous
distribution of outcomes. In addition, some random variables, such as height and
weight, are essentially continuous and have continuous distributions. In these
situations, the probability of a single event is not small as was assumed for the
discrete Poisson and normal distributions, and the equation = x does not
apply. The continuous normal distribution is given by:
1 xx

1
e 2
P( x) =
2
(5.35)
The properties (Fig.5.5) of the continuous normal distribution are:
P(x)
It is a continuous, symmetrical curve with both tails extending to infinity.

All three measures of central tendency, mean, median and mode, are
identical.
It is described by two parameters: the arithmetic mean x and the standard
deviation .
The mean x determines the location of the centre of the curve and the
standard deviation represents the spread around the mean.
Number of standard deviations

FIG.5.5. The continuous normal distribution indicating the probability levels at different
standard deviations (SDs) from the mean.
166
Relative response
Distance (mm)
FIG.5.6. Line source response curve obtained from a scintillation camera fitted to a normal
distribution model. Image resolution is measured as the distance of the full width at half
maximum (FWHM) of the percentage response. The standard deviation (SD) is the half width
at a percentage response of 60.65%.
All continuous normal distributions have the property that between the
mean and one standard deviation 68% is included on either side, between the
mean and two standard deviations 95%, and between the mean and three standard
deviations 99.7% of the total area under the curve.
5.3.4.2. Continuous normal distribution: applications in medical physics
The normal distribution is often used in radionuclide measurements and
imaging to fit to experimental data. In this case, the equation is modified as
follows:
P( x) = 100e
2
1 x x

(5.36)
where the maximum value of the distribution at x is normalized to 100.

The spatial resolution of imaging devices such as scintillation cameras and
positron emission tomography equipment is determined as the full width at half
maximum (FWHM) response of a normal distribution fitted to a point or line
spread function (Fig.5.6). The FWHM of the imaging device used in Fig.5.6
167
CHAPTER 5
was 23.6mm. The relation for a normal distribution between the FWHM and
standard deviation can be derived by setting P(x)=50 and solving Eq.(5.36):
FWHM=2.355 (5.37)
For the imaging system used in Fig.5.6, the standard deviation =10mm.
The value of the response P(x) is 60.65% at a distance of = x (Eq.(5.36)).
The value of the standard deviation can, therefore, also be obtained from the
measured percentage response curve by finding the x value at a percentage
response of 60.65% (Fig.5.6).
In radionuclide energy spectroscopy, the photopeak distribution can be
fitted to a normal distribution (Fig.5.1). The energy resolution of scintillation
detectors is expressed as the FWHM of the photopeak distribution divided by
the photopeak energy E. The energy spectrum in medical physics applications
is measured in kiloelectronvolts or megaelectronvolts. The fractional energy
resolution RE is:
RE =
FWHM 2.355
=
E
E
(5.38)
5.4. ESTIMATION OF THE PRECISION OF A SINGLE

MEASUREMENT IN SAMPLE COUNTING AND IMAGING
5.4.1. Assumption
A valuable application of counting statistics applies to the case in which
only a single measurement of a particular quantity is available and the uncertainty
associated with that measurement is required. The square root of the sample
variance should be a measure of the deviation of any one measurement from
the true mean value and will serve as an index of the degree of precision that
should be associated with a measurement from that set.
As only a single measurement is available, the sample variance cannot be
calculated directly using Eqs(5.3) or (5.9) and must be estimated by analogy
with an appropriate statistical model. The appropriate theoretical distribution
can be matched to the available data if the measurement has been drawn from
a population whose theoretical distribution function is predicted by either a
Poisson or Gaussian distribution. As the value of the single measurement x is the
168
only information available, it is assumed that the mean of the distribution is equal
to the single measurement:
xx
(5.39)
Having obtained an assumed value for x , the entire predicted probability

distribution function P(x) is defined for all values of x.
The expected sample variance s2 can be expressed in terms of the variance
of the selected statistical model:
s 2 = 2 = x x (5.40)
Therefore, the best estimate of the deviation from the true mean, which
should typify a single measurement x, is given by:
s = x (5.41)
To illustrate the application of Eq.(5.41), it is assumed that the probability

distribution function is Gaussian with a large value for the measurement x. The
range of values x or x x will contain the true mean with 68% probability.
If it is assumed that there is a single measurement x=100, then:
x = 100 = 10
In Table5.3, the various options available in quoting the uncertainty to be

associated with the single measurement are shown. The conventional choice is
to quote the measurement x plus orminus the standard deviation or 10010.
This interval is expected to contain the true mean x with a probability of 68%.
The probability that the true mean is included in the range can be increased by
expanding the interval associated with the measurement as is shown in Table5.3.
For example, to achieve a 99% probability that the true mean is included, the
interval must be expanded by 2.58. In the example, the range is then 10025.8.
When errors are reported, the associated probability level should be stated
in the report under methods.
169
CHAPTER 5
Table5.3. Examples of Error Intervals for a Single

Measurement x=100
Interval
(values)
Probability that the true mean
x0.67
93.3106.7
50
x1.00
90.0110.0
68
x1.64
83.6116.4
90
x2.00
80.0120.0
95
x2.58
74.2125.8
99
x3.00
70.0130.0
99.7
Interval
(relative )
x is included (%)
5.4.2. The importance of the fractional F as an indicator of the precision

of a single measurement in sample counting and imaging
The relation between the precision and a single counting measurement
x is given by Eq.(5.40). The precision, expressed as the standard deviation ,
will increase proportionally to the square root of the measurement x. Thus, if
the value of the single measurement x increases, the standard deviation will
also increase. The increase in the standard deviation will be smaller than that
of the measurement x. The relation between the standard deviation and the
single measurement is best demonstrated by calculating the relative or fractional
standard deviation F:
F =
x
1
=
=
(5.42)
x
x
x
Thus, the recorded number of counts or the value of the single measurement
x completely determines the relative precision. The relative precision decreases
as the number of counts increases. Therefore, to achieve a required relative
precision, aminimum number of counts must be accumulated.
The following example illustrates the important relation between the
relative precision and the number of counts recorded. If 100 counts are recorded,
the relative standard deviation is 10%. If 10 000 counts are recorded, the relative
standard deviation reduces to 1%. This example demonstrates the importance of
acquiring enough counts to meet the required precision.
It is easier to achieve the required precision when samples in counting tubes
are measured than when in vivo measurements on patients are performed. The
170
single measurement from a high count rate radioactive sample in a counting tube
will be obtained in a short time. However, if a low activity sample is measured,
the measurement time will have to be increased to achieve the desired precision.
The desired precision can be conveniently obtained by using automatic sample
counters. These counters can be set to stop counting after a preset time or preset
counts have been reached. By choosing the preset count option, the desired
precision can be achieved for each sample.
The acquisition time of in vivo measurements using collimated detector
probes, such as thyroid iodine uptake studies or imaging studies, can often not
be increased to achieve the desired precision as a result of patient movement. In
these single measurements, high sensitivity radiation detectors or a higher, but
acceptable radioactive dose, can be selected.
The precision of a single measurement is very important during radionuclide
imaging. If the number of counts acquired in a picture element or pixel is low, a
low precision is obtained. There will then be a wide range of fluctuations between
adjacent pixels. As a result of the poor quality of the images, it would only be
possible to identify large defect volumes or defects with a high contrast. To detect
a defect, the measured counts from the defect must lie outside the range of the
background measurement plus orminus two standard deviations (x2). During
imaging, the number of counts measured in a target volume will be determined
by the acquisition time, activity within the target volume and the sensitivity of
the measuring equipment. The sensitivity of imaging equipment can be increased
by increasing the FWHM spatial resolution. There is a trade-off between single
sample counting precision and the spatial resolution of the imaging device to
obtain images that would provide the maximum diagnostic value during visual
interpretation of the images by nuclear medicine physicians.
Counting statistics are also very important during image quantification
such as measuring renal function, left ventricular ejection fraction and tumour
uptake. During quantification, the accumulated counts by an organ or within a
target volume have to be accurately determined. In quantification studies, the
background activity, attenuation and scatter contributions have to be corrected.
These procedures further reduce the precision of quantification.
5.4.3. Caution on the use of the estimate of the precision of a
single measurement in sample counting and imaging
All conclusions are based on the measurement of a counted number of
success (number of heads in coin tossing). In nuclear measurements or imaging,
the estimate of the precision of a single measurement by using = x can only
be applied if x represents a counted number of success, that is the number of
events recorded in a given observation time.
171
CHAPTER 5
The estimate of the precision of a single measurement by using = x

cannot be used if x is not a directly measured count. For example, the association
does not apply to:
Counting rates;
Sums or differences of counts;
Averages of independent counts;
Pixel counts following tomographic image reconstruction;
Any derived quantity.
In these cases, the quantity is calculated as a function of the number of
counts recorded. The error to be associated with that quantity must be calculated
according to the error propagation methods outlined in the next section.
5.5. PROPAGATION OF ERROR
The preceding section described methods for estimating random error or
the precision of a single measurement during nuclear measurements or imaging.
Most procedures in nuclear medicine involve multiple nuclear measurements
and imaging procedures for the calculation of results such as thyroid iodine
uptake, ejection fraction, renal clearance, blood volume or red cell survival time,
on which clinical diagnosis is based. Similarly, internal dosimetry is performed
using nuclear measurements and imaging data. To estimate the corresponding
precision in the derived quantity, how the error associated with the initial
measurements propagates through the calculations that were performed to arrive
at the required result has to be followed. This is done by applying the error of
propagation formulas. The variables used in the calculation of errors must be
independent to avoid effects of correlation. It is assumed that the error in nuclear
measurements arises only from random fluctuations in the decay rate and is
statistically independent of other errors.
The error of propagation formulas applies to measurements that are
obtained from a continuous distribution as well as to Poisson and discrete
normal distributions. The measurements from continuous distributions will be
represented by x1, x2, x3... with variances of (x1)2, (x2)2, (x3)2... These equations
can be used to estimate precision in measurements such as height and weight.
Discrete nuclear measurements with Poisson or normal distribution
are represented by N1, N2, N3... with variances of (N1)2, (N2)2, (N3)2... or
N1, N2, N3...
172
5.5.1. Sums and differences

The product xs of the sums or difference of a series of measurements with a
continuous normal distribution is given by:
xs=x1x2x3
(5.43)
The variance of xs is given by:

(x1x2x3)2=(x1)2 + (x2)2 + (x3)2
(5.44)
The standard deviation is given by:

( x 1 x 2 x 3 ...) = ( x 1 ) 2 + ( x 2 ) 2 + ( x 3 ) 2 ... (5.45)
The fractional standard deviation is given by:

F ( x 1 x 2 x 3 ...) =
( x 1 ) 2 + ( x 2 ) 2 + ( x 3 ) 2 ...
x 1 x 2 x 3 ...
(5.46)
For counting measurements or measurements with a Poisson or discreet

normal distribution, the variance is given by:
( N 1 N 2 N 3 ...) 2 = N 1 N 2 N 3 ... (5.47)

( N 1 N 2 N 3 ...) = N 1 + N 2 + N 3 ... (5.48)

F ( N 1 N 2 N 3 ...) =
N 1 + N 2 + N 3 ...
N 1 N 2 N 3 ...
(5.49)
These equations apply to mixed combinations of sums and differences.
173
CHAPTER 5
TABLE5.4. Uncertainty after Summing and Subtracting

Counts
N2 N1
N2 N1
N1
500
22.4
0.0447
500
22.4
0.0447
N2
10
3.2
0.3162
450
21.2
0.0471
N1 N2
490
22.6
0.0461
50
30.8
0.6164
N1 + N2
510
22.6
0.0443
950
30.8
0.0324
The influence on the standard deviation and fractional standard deviation

of summing and subtracting values N1 and N2 is demonstrated in Table5.4. The
following conclusions can be drawn:
The standard deviation for N1 N2 and N1 + N2 is the same for the same
values of N1 and N2, but the fractional standard deviation F is different;
The fractional standard deviation for differences is large when the
differences between the values are small.
This is the reason why it is important to limit the background to a value as
low as possible in counting procedures. In imaging, when scatter or background
correction is performed by subtraction, image quality deteriorates as a result of
the increased uncertainty in the pixel values.
5.5.2. Multiplication and division by a constant
We define:
xM=Ax (5.50)
where A is a constant.
Then:
M=Ax (5.51)
and
174
A x x
(5.52)
=
Ax
x
F =

normal distribution, the following applies:
xM=AN (5.53)
Then:
M = A N (5.54)
and
1
(5.55)
N
F =
Similarly, if:
x
(5.56)
B
xD =
where B is also a constant:

M =
x
(5.57)
B
and
F =
x B x
=
(5.58)
B x
x

normal distribution, the following apply:
xD =
N
(5.59)
B
M =
N
(5.60)
B
175
CHAPTER 5
and
1
(5.61)
N
F =
It should be noted that multiplying (Eqs(5.52) and (5.55)) or dividing

(Eqs(5.58) and (5.61)) a value by a constant does not change the fractional
standard deviation.
5.5.3. Products and ratios
The uncertainty in the product or ratio of a series of measurements
x1, x2, x3... is expressed in terms of the fractional uncertainties in the individual
results, F(x1), F(x2), F(x3)...
The product xP of the products or ratios of a series of measurements with a
continuous normal distribution is given by:
x P = x1
x2
x3
... ... ... ... ... (5.62)
The notation means x 1 x 2 or x 1 x 2 . These equations apply to mixed

combinations of sums and differences.
The fractional variance of xP is given by:
F (x1
x2
x3
... ...) 2 = F ( x 1 ) 2 + F ( x 2 ) 2 + F ( x 3 ) 2 ... ... (5.63)

F (x1
x2
x3
... ...) = F ( x 1 ) 2 + F ( x 2 ) 2 + F ( x 3 ) 2 ... ... (5.64)

( x 1
x2
x3
... ) = F ( x 1 ) 2 + F ( x 2 ) 2 + F ( x 3 ) 2 ... ( x 1
x2
x3
... ) (5.65)

normal distribution, the product or ratio is given by:
NP = N1
N2
N3
... ... (5.66)
The fractional variance of NP is given by:

F (N 1
176
N2
N3
... ...) 2 =
1
1
1
(5.67)
+
+
+ ... ...
N1 N2 N3

F (N 1
N2
N3
... ...) =
1
1
1
+
+
+ ... ... (5.68)
N1 N2 N3

( N 1
N2
N3
... ...) =
1
1
1
+
+
+ ... ... ( N 1
N1 N2 N3
N2
N3
... ...) (5.69)
5.6. APPLICATIONS OF STATISTICAL ANALYSIS

5.6.1. Multiple independent counts
5.6.1.1. Sum of multiple independent counts
If it is supposed that there are n repeated counts from the same source for
equal counting times and the results of the measurements are N1, N2, N3....... Nn
and their sum is Ns, then:
N s = N 1 + N 2 + N 3 ... (5.70)
According to the propagation of error for sums and Eq.(5.48):

N = N 1 + N 2 + N 3 ... = N s (5.71)
s
The results show that the standard deviation for the sum of all counts is the
same as if the measurement had been carried out by performing a single count,
extending over the period represented by all of the counts.
5.6.1.2. Mean value of multiple independent counts
If the mean value N of the n independent counts referred to in the previous
section is calculated, then:
N=
Ns
(5.72)
n
177
CHAPTER 5
Equation (5.72) is an example of dividing an error-associated quantity N

by a constant n. Equation (5.51), therefore, applies and the standard deviation of
the mean or standard error is given by:
N =
N
n
Ns
n
nN
N
(5.73)
=
n
n
It should be noted that the standard deviation for a single measurement Ni

(Eq.(5.41)) is N i = N i .
A typical count will not differ greatly from the mean N i N . Thus, the
mean value based on n independent counts will have an expected error that is
smaller by a factor of n compared with any single measurement on which the
mean is based. To improve the statistical precision of a given measurement by a
factor of two, the counting time must, therefore, be increased four times.
5.6.2. Standard deviation and relative standard
deviation for counting rates
If N counts are accumulated over time t, then the counting rate R is given
by:
R=
N
(5.74)
t
In the above equation, it is assumed that the time t is measured with a very
small uncertainty, so that t can be considered a constant. The calculation of the
uncertainty associated with the counting rate is an application of the propagation
of errors, multiplying by a constant (Eq.(5.60)):
R =
x
N
R
=
=
(5.75)
t
t
t
The fractional standard deviation is calculated using Eq.(5.61):

F =
x
N
1
=
=
(5.76)
tR
tR
tR
The above equations illustrate the calculation of uncertainties if calculations

are required to obtain a value, and the equation for a single value (Section 5.3)
cannot be applied. The following example illustrates the use of the equations.
178
5.6.2.1. Example: comparison of error of count rates and counts accumulated

The activity of two samples is measured. Sample 1 is counted with a
counter that is set to stop when a count of 10000 is reached. It takes 100 s to
reach 10000 counts. Sample 2 is counted using an automatic sample changer.
The activity of the sample is given as 10000 counts per second (cps) and the
sample was counted for 100 s.
Calculating the counting error associated with the measurements of
samples 1 and 2:
Sample 1:
The counts acquired: N=10 000 counts
Standard deviation (Eq.(5.41)): s N = 100 counts
Fractional standard deviation (Eq.(5.42)): F = 0.01 = 1%
Sample 2:
The count rate: 10000 cps
10 000
= 10 cps
Standard deviation (Eq.(5.75)): R =
100
Fractional standard deviation (Eq.(5.76)):

F =
1
= 0.001 = 0.1%
10 000 100
Although the counts acquired for sample 1 and the count rate of sample 2
were numerically the same, the uncertainties associated with the measurements
were very different. When calculations on counts are performed, it must be
determined whether the value is a single value or whether it is a value that has
been obtained by calculation.
5.6.3. Effects of background counts
Background counts are those counts that do not originate from the sample
or target volume or are unwanted counts such as scatter. The background
counts during sample counting consist of electronic noise, detection of cosmic
rays, natural radioactivity in the detector, and down scatter radioactivity from
non-target radionuclides in the sample. During in vivo measurements, such
as measurement of thyroid iodine uptake or left ventricular ejection fraction,
radiation from non-target tissue will also contribute to background. Scattered
179
CHAPTER 5
radiation from target as well as non-target tissue will influence quantification and
will be included in the background. To obtain the true net counts, the background
is subtracted from the gross counts accumulated. The uncertainty of the true
target counts can be calculated using Eqs(5.48) and (5.49), and the uncertainty of
true count rates can be calculated using Eqs(5.75) and (5.76).
If the background count is Nb, and the gross counts of the sample and
background is Ng, then the net sample count Ns is:
N s = N g N b (5.77)
The standard deviation for Ns counts is given by Eq.(5.48):

( N s ) = N g + N b (5.78)
The fractional standard deviation for Ns counts is given by Eq.(5.49):

F (N s ) =
Ng + Nb
Ng Nb
(5.79)
If the background count rate is Rb, acquired in time tb, and the gross count
rate of the sample and background is Rg, acquired in time tg, then the net sample
count rate Rs is:
Rs = Rg R b (5.80)
The standard deviation for a count rate Rs is given by Eqs (5.45) and(5.75):
(Rs ) =
Rg
tg
Rb
(5.81)
tb
The fractional standard deviation for a count rate Rs is given by Eqs (5.46)
and(5.76):
Rg
F (Rs ) =
tg
Rb
tb
Rg R b
(5.82)
If the same counting time t is used for both sample and background
measurement:
(R s ) =
180
Rg + R b
t
(5.83)
and
F (R s ) =
Rg + R b
t (R g R b )
(5.84)
5.6.3.1. Example: error in net target counts following background correction

The following example illustrates the application to determine the
uncertainty in the measurement of target volume counts following background
correction. A planar image of the liver is acquired for the detection of tumours.
Two equal sized ROIs, ROI1 and ROI2, were selected to cover the areas of the
two potential tumours. The gross counts Ng in ROI1 were 484 counts (Table5.5)
and in ROI2 484 counts. The background counts Nb selected over normal tissue of
the same area as for the gross counts were 441 and 169 counts. How to calculate
the uncertainties in the tumor volume net counts is presented.
The difference and error associated with the difference (Eq.(5.77)
Eq. (5.79)) when Ng Nb are:
Ng Nb=484 441=43 counts
( N g N b ) = 484 + 441 = 30.4 counts
F (N g N b ) =
484 + 441
= 0.7073
484 441
P ( N g N b ) = 70.7%
The influence on the standard deviation and fractional standard deviation of

background correction for Ng Nb and Ng Nb is demonstrated in Table5.5. The
following conclusion can be drawn: the fractional F and percentage P standard
deviations significantly increase when the background increases relative to the
net counts.
This is the reason why it is important in measurements of radioactivity to
acquire as many counts as possible to decrease the uncertainty in detection of
target volume radioactivity. The following example illustrates the application
to determine the uncertainty in the measurement of target volume count rate
following background correction.
181
CHAPTER 5
TABLE5.5. CALCULATION OF UNCERTAINTIES IN COUNTS AS A

RESULT OF BACKGROUND CORRECTION
Ng Nb
Source
Counts counts
Ng Nb
P (%) Source Counts counts
P (%)
Ng
484
22.0
0.0455
4.5
Ng
484
22.0
0.0455
4.5
Nb
441
21.0
0.0476
4.8
Nb
169
13.0
0.0769
7.7
Ns
43
30.4
0.7073
70.7
Ns
315
25.6
0.0811
8.1
3 (Ns)
91
Counts
Not
significant
3 (Ns)
77
Counts Significant
5.6.3.2. Example: error in net target count rate following background correction
A planar image of the liver is acquired for the detection of tumours.
Two equal sized ROIs, ROI1 and ROI2, were selected to cover the areas of
the two potential tumours. The gross count rate Rg in ROI1 was 484 counts
perminute (cpm) (Table5.6) and in ROI2 484 cpm. The background count rates
Rb selected over normal tissue of the same area as for the gross counts were 441
and 169 cpm. The acquisition time of the image was 2min. How to calculate the
uncertainties in the tumor volume net counts is presented.
The difference and error associated with the difference (Eq.(5.80)
Eq.(5.82)) when Rg Rb are:
Rg R b = 484 441 = 43 cpm
(R g R b ) =
484 441
+
= 21.5 cpm
2
2
484 441
+
2
2 = 0.5001
F (R g R b ) =
484 441
P (Rg R b ) = 50.0%
The influence on the standard deviation and fractional standard deviation

of background correction for Rg Rb and Rg Rg is demonstrated in Table5.6.
182
Again, it is shown that the fractional standard deviation F significantly increases

when the background count rate increases relative to the net target count rate.
TABLE5.6. Calculation of Uncertainties in Count Rates as A
Result of Background Correction
Rg Rb
Source
Count rate
(cpm)
(cpm)
P (%)
Source
Rg Rb
Count rate
F (cpm)
(cpm)
P (%)
Rg
484
15.6
0.0321
3.2
Rg
484
15.6
0.0321
3.2
Rb
441
14.8
0.0337
3.4
Rb
169
9.2
0.0544
5.4
Rs
43
21.5
0.5001
50.0
Rs
315
18.1
0.0574
5.7
Minutes
Minutes
3 (Rs)
65
Not
significant
3 (Rs)
54
Significant
5.6.4. Significance of differences between counting measurements

If N1 and N2 counts are measured in two counting measurements, the
difference (N1 N2) between the measured counts may be a result of random
variations in the counting rate or may be as a result of an actual difference.
The statistical significance of the difference is evaluated by comparing it to the
expected random error expressed as the standard deviation d of the difference.
If (N1 N2) > 2(N1 N2), there is a 5% chance that the difference is caused by
random error (seeTable5.3). If:
N 1 N 2 > 3( N 1 N 2 ) (5.85)
there is a 0.3% chance that the difference is caused by random error and this
difference is considered significant.
The examples in the previous section to determine whether tumours
were present following a liver scan illustrate the application to determine the
significance of the difference between two counts (Table5.5). The net counts
and uncertainty over two tumour areas were calculated. Do the counts over the
tumour areas significantly differ from the normal background area?
For the difference for Ng Nb (Table5.5) to be significant, Eq.(5.85) must
apply.
183
CHAPTER 5
The difference of 43 cpm was less than the norm of 3(N1 N2) and the
difference is, therefore, not significant. It can be concluded with a smaller than
0.3% chance that there is not a tumour present.
An example when Ng Nb is also given in Table5.5. In this case, the
315 cpm counts difference was larger than 3(N1 N2) of 77 cpm. The difference
in this case is significant. It can be concluded with a smaller than 0.3% chance
that there is a tumour present.
The significance of differences between the counting rates of samples can
also be calculated. Two counting rates, R1 and R2, are acquired using counting
times t1 and t2.
The uncertainty associated with the difference is given by applying
Eqs(5.45) and (5.75):
( R1 R 2 ) =
R1 R 2
(5.86)
+
t1
t2
For the difference R1 R2 to be significant:

R1 R2 > 3(R1 R2) (5.87)
The examples in the previous section (Table5.6) to determine whether
tumours were present following a liver scan illustrate an application to determine
the significance of the difference between two count rates. The net count rate and
uncertainty over two tumour areas were calculated. Do the count rates over the
tumour areas significantly differ from the normal background area?
For the difference for Rg Rb (Table5.6) to be significant, Eq.(5.87) must
apply.
The difference count rate of 43 cpm was less than the 65 cpm which is the
norm of 3(R1 R 2 ) and the difference is, therefore, not significant. It can be
concluded with a smaller than 0.3% chance that there is not a tumour present.
An example when Rg Rb is also given in Table5.6. In this case, the
difference of 315 cpm was larger than 3(R1 R 2 ) which was 54 cpm. The
difference in this case is significant. It can be concluded with a smaller than 0.3%
chance that there is a tumour present.
5.6.5. Minimum detectable counts, count rate and activity
According to Eq.(5.85), if the difference of two measurements is larger
than three standard deviations, the difference is considered significant. Therefore,
184
theminimum net counts Nm that can be detected with 0.3% confidence is given
by:
Nm=N1 N2=3(N1 N2) (5.88)
or
Nm=Ng Nb=3(Ng Nb) (5.89)
Solving this equation for Ng will give theminimum detectable gross
counts Nm:
(2 N b + 9) + 72 N b + 81
Ng =
(5. 90)
An approximation can be used by assuming that Ng Nb and:

N g N b + 3 2 N b (5.91)
Theminimum detectable activity Am can be calculated:

Am =
Nm
(5.92)
tS
where
S is the sensitivity of the detection system usually expressed as count rate per
becquerel;
and t is the time that the background was counted.
5.6.5.1. Example: calculation ofminimum activity that can be detected
A detector is to be used to detect 131I in the thyroid of radiation workers.
The background count was 441 counts measured over a period of 5min. The
acquisition time for the thyroid was also 5min. The sensitivity of the counter was
0.1 counts s1 Bq1. What is theminimum activity that can be detected?
From Eq.(5.90):
Ng =
(2 N b + 9) + 72 N b + 81
2
(2 441 + 9) + 72 441 + 81
= 535 counts
2
185
CHAPTER 5
It should be noted that Ng Nb=94 counts and 3(Ng Nb)=94 as was

specified in Eq.(5.85). Theminimum detectable radioactivity is:
Am =
(535 441)
= 3.124 Bq
5 60 0.1
Theminimum detectable net count rate Rm is given by Eq.(5.89):

Rm = Rg R b > 3(Rg R b ) (5.93)
Solving this equation for Rg gives theminimum detectable gross count

rate Rm:
36 R b 81 36 R b
9
+ 2+
2 R b + +
t g
tg
tb
tg
Rg =

2
(5.94)
An approximation can be used by assuming that Rg Rb and from Eqs(5.86)

and (5.87):
Rg R b + 3
Rb Rb
+
(5.95)
tg
tb
Theminimum detectable activity Am can be calculated:

Am =
Rm
(5.96)
S
where S is the sensitivity of the detection system usually expressed as count rate
per becquerel.
5.6.5.2. Example 2: calculation ofminimum activity that can be detected
A detector is to be used to detect 131I in the thyroid of radiation workers.
The background count rate was 441 cpm measured over a period of 5min and the
thyroid count rate was measured over 1min. The sensitivity of the counter was
0.1 counts s1 Bq1. What is theminimum activity that can be detected?
From Eq.5.94:
2 441 + 9 + 36 441 + 81 + 36 441
1
1
5
12
Rg =
= 515 cpm
2
186
It should be noted that Rg Rb=74 cpm and 3(Rg Rb)=74 cpm as was
specified in Eq.(5.93). Theminimum detectable radioactivity is:
Am =
(515 441)
= 12.28 Bq
0.1 60
5.6.6. Comparing counting systems

It was concluded in Section 5.3.1 that a large number of counts have smaller
uncertainties expressed as the fractional standard deviation. In Section 5.6.3, it
was shown that if background counts increase, the uncertainty of the net counts
expressed as fractional standard deviation rapidly increases. Thus, it is desirable
to use a counting system with a high sensitivity and low background. However,
when the detector sensitivity is increased, the system will also be more sensitive
to background. The trade-off between sensitivity and background can be analysed
as follows.
It is considered that results from systems 1 and 2 are compared. The
acquisition times for gross and background counts are acquired over the same
time. From Eq.(5.79):
N g1 + N b1
F1( N S1 ) =
N g1 N b1
and
F2 ( N S2 ) =
N g2 + N b2
N g2 N b2
The fractional uncertainties for the net sample counts obtained with the two
systems are, therefore:
N g1 + N b1
N g1 N b1
F1( N S1 )
(5.97)
=
F2 ( N S2 )
N g2 + N b2
N g2 N b2
If
F1( N S1 )
< 1, then system 1 is statistically the preferred system. If
F2 ( N S2 )
F1( N S1 )
> 1 , then system 2 is preferred.
F2 ( N S2 )
187
CHAPTER 5
Systems can be compared using the count rate and fractional standard
deviation for the count rate RS (Eq.(5.82)). To compare systems 1 and 2, the ratio
of the fractional standard deviation is calculated:
Rg1
t g1
R b1
t b1
Rg1 R b1
F1(RS1 )
(5.98)
=
F2 (RS2 )
Rg2 R b2
+
t g2
t b2
Rg2 R b2
Equation (5.98) can be used to compare different counting times in the

same system for measuring fixed geometry samples. However, to obtain the
best energy window selection in a system, or to compare two systems, the same
counting time t should be used:
Rg1 + R b1
Rg1 R b1
F1(RS1 )
(5.99)
=
F2 (RS2 )
Rg2 + R b2
Rg2 R b2
It should be noted that Eqs(5.98) and (5.99) are the same except that in
Eq.(5.99) counts are substituted by counting rates.
Equation (5.99) can also be used in planar imaging. Different collimators
can be evaluated by comparing counts from a target region to a non-target or
background region. However, in imaging, spatial resolution is also important and
must be considered.
5.6.7. Estimating required counting times
It is supposed that it is desired to determine the net sample or target
counting rate Rs to within a certain fractional uncertainty F(Rs). It is supposed
further that the approximate gross sample Rga and background Rba counting rates
are known from preliminary measurements. If a counting time t is to be used
for both the sample or target and the background counting measurements, then
the time required to achieve the desired level of statistical reliability is given by
Eq.(5.84):
188
t=
Rga + R ba
2
[ F (Rs )](Rga R ba ) 2
5.6.7.1. Example: calculation of required counting time

The counting time for a thyroid uptake study using a collimated detector is
to be determined. The preliminary measurement of the gross thyroid count rate is
Rga=900 cpm and background count rate Rba=100 cpm. What counting time is
required to determine the net count rate to within 5%?
Rsa=900 100=800 cpm
t=
(900 + 100)
1000
=
= 0.625 min
2
2
(0.05) (900 100)
(0.05) 2 (800) 2
The time for both the thyroid and background counts is 0.625min, resulting
in a total time of 1.25min.
5.6.8. Calculating uncertainties in the measurement
of plasma volume in patients
A plasma volume (PV) measurement is required on a patient and the
uncertainty in the PV measurement is to be calculated. The PV is measured by
using the dilution principle. A labelled plasma sample of a known volume is
prepared for injection into the patient. A standard sample with the same activity
and volume is also prepared for counting. The standard sample is diluted before
a sample is counted. Tenminutes after injection of the sample, a blood sample
is obtained, the plasma separated from the blood and the blood sample counted.
The PV is calculated using the following equation:
PV =
Rs
VD (5.100)
Rp
where
Net count rate per millilitre of standard sample Rs=Rs+b Rb;
Rb is the count rate of background;
Rs+b is the gross count rate per millilitre of standard sample;
Net count rate per millilitre of plasma sample Rp=Rp+b Rb;
189
CHAPTER 5
Rp+b is gross count rate per millilitre of plasma sample;

V is volume of standard sample in millilitres with percentage uncertainty P(V);
and D is dilution of standard sample for counting with percentage
uncertainty P(D).
TABLE5.7. APPLICATION OF THE PROPAGATION OF ERRORS
PRINCIPLE TO THE CALCULATION OF UNCERTAINTIES
Values
Symbol
Value
Uncertainty in values
Unit
Symbol
Calculation
F(%)
10
min
Rs+b
3200
cpm
(Rs+b)
Rs+b
t
17.89 0.559
Rb
200
cpm
(Rb)
Rb
t
4.472 2.236
Rs
3000
cpm
(Rs)
((Rs+b )) 2 + ((R b )) 2
18.44 0.615
Rp+b
1200
cpm
(Rp+b)
Rp
1000
cpm
(Rp)
Rs
Rp
mL
(V)
Rs
V
Rp
15
mL
200
PV=
Rs
VD
Rp
(Rs/Rp)
R
s
V
R p
R p+b
10.95 0.913
((R p+b )) 2 + ((R b )) 2
Rs
Rp
(R ) 2 (R p )
R + R
s
p
0.040 1.333
0.150 3.000
R
s
V
R p
(R / R ) 2 (V ) 2
s
p
+ V
Rs / R p
(D)
0.492 3.283
6.000 3.000
3000
mL
(PV)
((R / R )V)
(D) 2
s
p
PV
+
(Rs / R p )V
D
The following values were used and measured:

Counting time t =10min
VP(V)
=53% mL
DP(D)
=2003%
Rs+b
=3200 cpm
Rp+b
=1200 cpm
Rb
=200 cpm
190
11.83 1.183
2
133
4.447
The uncertainties are calculated step by step by applying the propagation of

errors principle (seeTable5.7)
The measured PV is, therefore, 3000133 mL or 30004.447%. It
should be noted that the uncertainty is expressed as one standard deviation. A
spreadsheet can be used efficiently to do the calculations in the above table. With
a spreadsheet, the influence in changing the counting time or uncertainties in
the measurement of the dilution and volume of the standard can be investigated.
These spreadsheets are ideally suited for calculations of uncertainties in routine
clinical investigations.
5.7. APPLICATION OF STATISTICAL ANALYSIS:
DETECTOR PERFORMANCE
5.7.1. Energy resolution of scintillation detectors
We have directed our attention in the previous sections to determine the
uncertainty associated with the number of counts measured in a radioactive
sample or number of counts in an image pixel. Poisson statistics also play an
important role in other aspects of the detection of radiation. A statistical process
determines the energy resolution of a detector or the uncertainty associated
with the energy measurement of a detected photon. This is the reason why the
energy resolution of a solid state detector is significantly better than that of a
scintillation detector. The type of detector and the energy of the detected photons
determine the energy resolution or uncertainty in the energy of a detected photon.
The energy resolution for a detector system and a specific radionuclide does not
change from sample to sample. This is different from counting statistics where
the uncertainty is determined by the number of counts accumulated during a
measurement. Therefore, even for the same sample and same detector system, the
uncertainty can change if measurements are repeated following the decay of the
nuclide.
Another important consequence of statistics is that in scintillation cameras
the location of the position of incoming photons is based on the pulses detected
by the detectors. Therefore, the statistics of the detector system limits the spatial
resolution that can be achieved with an imaging device. A clear understanding of
the statistics associated with the detector when detecting a photon is, therefore,
important.
In this section, we will investigate the statistical processes in scintillation
detectors, since they are widely used in nuclear medicine for sample counting
and imaging.
191
CHAPTER 5
The operation of scintillation detectors can be considered a three stage

process:
(a) The number x of light photons produced in the scintillator by the detected
ray;
(b) The fraction p of the light photons that will eject electrons from the
photocathode of the PMT;
(c) The multiplication M of these electrons multiplied at successive dynodes
before being collected at the anode.
The average number Ne of electrons produced at the anode is given by:
N e = xpM (5.101)
The fractional variance F2 in the electron number N for a three stage

cascade process is given by Eq.(5.63):
F2 ( N e ) = F2 ( x) + F2 ( px) + F2 ( pxM ) (5.102)
It can be shown that for dynodes with identical multiplication
1
. It is assumed that the production of light photons follows
M 1
1
a Poisson distribution and, therefore, F2 ( x) = . The fractional variance of
x
F2 (M ) =
the production of electrons from light photons at the photocathode is given by

Eq.(5.20) as F2 ( p) =
F2 ( N e ) =
1 p
:
p
1 1 1 p
1
1
(5.103)
+
+
x x p
xp (M 1)
The fractional energy resolution RE of detectors is expressed as the FWHM

divided by the mean photon energy (Section 5.3.4.1). From Eq.(5.38):
RE =
FWHM 2.355(E) 2.355( N e )

=
=
(5.104)
E
E
Ne
From Eqs(5.103) and (5.104):

R E = 2.355
192
1
(5.105)
1 p
1
+
x 1 +
p
p(M 1)
5.7.2. Intervals between successive events

The time intervals separating random events are of interest in nuclear
measurements. Such an application is the calculation and measurement of the
paralysable dead time of counting systems.
If r is the average rate at which events are occurring, it follows that r dt is
the differential probability that an event will take place in the differential time
increment dt. For a radiation detector with unity efficiency, the time interval for
counting a single radionuclide is given by:
r=
dN
= lN
dt
where
N
is the number of radioactive nuclei;
and is their decay constant.

In order to derive a distribution function to describe the time interval
between adjacent random events, it is first assumed that an event has occurred at
time t=0. What is the differential probability that the next event will take place
within a differential time dt after a time interval t?
Two independent processes must take place: no events may occur within
the time interval from 0 to t, but an event must take place in the next differential
time increment dt. The overall probability will then be given by the product of
the probabilities characterizing the two processes, or:
Probability of next event Probability of number
taking place in dt after
= of events during time
delay of t
from 0 to t
P1(t ) dt
= P(0)
Probability of
event during dt
r dt
(5.106)
The first factor on the right hand side follows directly from the earlier
discussion of the Poisson distribution. We seek the possibility that no events will
be recorded over an interval of length t for which the average number of recorded
events should be rt. From Eq.(5.23):
P(0) =
(rt ) 0 e rt
0!
193
CHAPTER 5
P(0) = e rt (5.107)
Substituting Eq.(5.107) into Eq.(5.106):

P1(t ) dt = re rt dt (5.108)
P1(t) is now the distribution function for intervals between adjacent random
events. Figure5.7 shows the simple exponential shape of the distribution.
FIG.5.7. Distribution for intervals between adjacent random events.
It should be noted that the most probable distribution is zero. The average
interval length is calculated by applying Eq.(5.8):
t =
=
P (t ) d t
tP1(t ) dt
1
te rt dt
=
e
rt
dt
1
(5.109)
r
5.7.3. Paralysable dead time

In the paralysable dead time model, a fixed dead time follows each event
during the live period of the detector. However, events that occur during the
194
dead period, although not recorded, still create another fixed dead time on the
system following the lost event. The recorded rate of events m is identical to the
rate of occurrences of time intervals between true events, which exceed . The
probability of intervals larger than can be obtained by integrating Eq.(5.108):
P2 (t ) dt =
P1(t ) dt = e r (5.110)
The rate occurrence m of such intervals is obtained by multiplying

Eq.(5.110) by the true rate r:
m = re r
(5.111)
There is no explicit solution for r; it must be solved iteratively to calculate r

from measurements of m and . This can be done using a spreadsheet.
Bibliography
BUSHBERG, J.T., SEIBERT, J.A., LEIDHOLDT, E.M., BOONE, J.M., The Essential Physics
of Medical Imaging, Lippincott Williams and Wilkins, London (2002).
CHERRY, S.R., SORENSON, J.A., PHELPS, M.E., Physics in Nuclear Medicine, Saunders,
Los Angeles, CA (2003).
DELANEY, C.F.G., FINCH, E.C., Radiation Detectors, Clarendon Press, Oxford (1992).
KNOLL, G.F., Radiation Detection and Measurement, John Wiley and Sons, New York (1989).
NATIONAL ELECTRICAL MANUFACTURERS ASSOCIATION, Standards Publication
NU 1-2007, Performance Measurements of Gamma Cameras (2007).
195
CHAPTER 6
BASIC RADIATION DETECTORS
C.W.E. VAN EIJK
Faculty of Applied Sciences,
Delft University of Technology,
Delft, Netherlands
6.1. INTRODUCTION
6.1.1. Radiation detectors complexity and relevance
Radiation detectors are of paramount importance in nuclear medicine. The
detectors provide a wide range of information including the radiation dose of
a laboratory worker and the positron emission tomography (PET) image of a
patient. Consequently, detectors with strongly differing specifications are used.
In this chapter, general aspects of detectors are discussed.
6.1.2. Interaction mechanisms, signal formation and detector type
A radiation detector is a sensor that upon interaction with radiation
produces a signal that can preferably be processed electronically to give the
requested information. The interaction mechanisms for X rays and rays are the
photoelectric effect, Compton scattering and pair formation, where the relative
importance depends on the radiation energy and the interaction medium. These
processes result in the production of energetic electrons which eventually
transfer their energy to the interaction medium by ionization and excitation.
Charged particles, such as particles, transfer their energy directly by ionization
and excitation. In all cases, the ionization results either in the production of
charge carriers, viz. electrons and ions in a gaseous detection medium, and
electrons and holes in a semiconductor detector material, or in the emission of
light quanta in a scintillator. These processes represent the three major groups
of radiation detectors, i.e. gas filled, semiconductor and scintillation detectors.
In the former two cases, a signal, charge or current is obtained from the
detector as a consequence of the motion of charge in the applied electric field
(Figs6.1(a)and (b)). In the scintillation detector, light emission is observed by
means of a light sensor that produces observable charge or current (Fig.6.1(c)). A
detailed discussion is presented in Sections 6.26.4.
196
6.1.3. Counting, current, integrating mode

In radiology and radiotherapy, radiation detectors are operated in current
mode. The intensities are too high for individual counting of events. In nuclear
medicine, on the contrary, counting mode is primarily used. Observing individual
events has the advantage that energy and arrival time information are obtained,
which would be lost in current mode. In the case of a personal dosimeter, the
detector is used in integrating mode. The dose is, for example, measured monthly.
Furthermore, instead of real time observation, the information is extracted at a
much later time after the actual interaction.
Electrons
Electrons
Light quanta
Electron
from
interaction
Ions
Gas
Holes
Scintillator
Light sensor
FIG.6.1. Principle of operation of (a) a gas filled detector, i.e. an ionization chamber;
(b)a semiconductor detector, i.e. a silicon detector; and (c)a scintillation detector. The former
two detectors are capacitors. The motion of charge results in an observable signal. The light of
a scintillation detector is usually detected by a photomultiplier tube.
6.1.4. Detector requirements

The quality of a radiation detector is expressed in terms of sensitivity,
energy, time and position resolution, and the counting rate a detector can handle.
Obviously, other aspects such as cost, machinability and reliability are also very
important. The latter will not be discussed in this chapter.
6.1.4.1. Sensitivity
In radiation detection, the sensitivity depends on (i) the solid angle
subtended by the detector and (ii) the efficiency of the detector for interaction
with the radiation. The first point will be obvious and is not discussed further.
In nuclear medicine, relevant X ray and ray energies are in the range of
~30511keV. The detection efficiency is governed by the photoelectric effect
and Compton scattering only. The attenuation length (in centimetres) of the
197
CHAPTER 6
former is proportional to Zeff34, where is the density and Zeff is the effective
atomic number of the compound. Compton scattering is almost independent of Z;
it is just proportional to . The density of a gas filled detector is three orders of
magnitude smaller than that of a solid state detector. Thus, solid state detectors
are very important in nuclear medicine. At 511keV, even the highest possible
and Zeff are needed. Gas filled detectors are used in dosimetry.
6.1.4.2. Energy, time and position resolution
Energy, time and position resolution depend on a number of factors. These
are different depending on the physical property considered and the type of
detector; yet, there is one aspect in common. Resolution is strongly coupled to
the statistics of the number of information carriers. For radiation energy E, this
number is given by N=E/W in which W is the mean energy needed to produce
an information carrier. Typical W values are shown in Table6.1. As the smallest
number of information carriers in the process of signal formation is determinative,
for scintillation the effect of the light sensor is also shown. From the W values,
it can be seen that semiconductor detectors produce the largest number of
information carriers and inorganic scintillators coupled to a photomultiplier tube
(PMT) the smallest. If a ray energy spectrum is measured, the observed energy
resolution is defined as the width of a line at half height (FWHM: full width at
half maximum) E divided by its energy E. With N =E/W, and N being the
corresponding FWHM:
E N
FW
=
= 2.35
(6.1)
E
N
E
where
N is 2.35 for a Gaussian distribution;
2 is FN for the variance;
and F is the Fano factor. For gas filled detectors, F=0.050.20, for
semiconductors F 0.12. For a scintillator, F=1.
Using the corresponding F and W values, it can be seen from Eq.(6.1)
that the energy resolution of a semiconductor is ~16 times higher than that
of an inorganic scintillator PMT. In this discussion, other contributions to the
energy resolution were neglected, viz. from electronic noise in the case of the
semiconductor detector and from scintillator and PMT related effects in the
198
other case. Nevertheless, the large difference, by an order of magnitude, is

characteristic of the energy resolutions.
TABLE6.1. MEAN
CARRIERS
ENERGIES
W TO
Detector type
Gas filled (electronion)
Semiconductor (electronhole)
Inorganic scintillator (light quantum)
Inorganic scintillator + photomultiplier tube (electron)
Inorganic scintillator + silicon diode (electronhole pair)
PRODUCE
INFORMATION
W (eV)
30
3
25
100
35
In nuclear medicine, time resolution is mainly of importance for PET. Time

resolution depends primarily on two factors, the rise time and the height of the
signal pulses. The effect of the former can be understood by considering that it
is easier to measure the position of a pulse on a timescale with an accuracy of
100 ps if the rise time is 1 ns, than if it is 10 ns. The pulse height is important
because there is noise as well. The higher the pulse relative to the noise, the
easier it is to determine its position. In addition, time jitter due to pulse height
(energy) variation will become less important. If time resolution is the issue, the
fast response and fast rise time of inorganic scintillators and the fast response of
the light sensors make the scintillator the preferred detector.
Position resolution can be obtained easiest by pixelating the detector at a
pitch corresponding to the requested resolution. In nuclear medicine, position
resolution is an issue in ray detection in the gamma camera and in single photon
emission computed tomography (SPECT) and PET detection systems. In the
latter, pixelated scintillators are used and the position resolution of a detector
is determined by the pitch. More recently, studies have been published on the
use of monolithic scintillator blocks in PET. Light detection occurs by means
of pixelated sensors. In principle, this is analogous to the gamma camera. A
relatively broad light distribution is measured using pixels that are smaller in size
to define the centre of the distribution, thus obtaining a position resolution that is
even better than the pixel size.
6.1.4.3. Counting rate and dead time
An achievable counting rate depends on (i) the response time of a detector,
i.e. the time it takes to transport the charge carriers to form the signal or to emit
199
CHAPTER 6
the scintillation light, and (ii) the time needed to process the signals and to handle
the data. For a better understanding, the concept of dead time is introduced.
It is theminimum time separation between interactions (true events) at which
these are counted separately. Non-paralysable and paralysable dead time are
considered. In the former case, if within a period of time after a true event a
second true event occurs, it cannot be observed. If the second event occurs at a
time t > , it will be counted. The dead period is of fixed length . Defining true
event rate T (number per unit time) and counting rate R, the fraction of time the
system is dead is given by R and the rate of loss of events is TR. Considering
that the latter is also T R, the non-paralysable case can be derived:
R=
T
(6.2)
1+T
If in the paralysable model a second event occurs at t > after the first
event, it will be counted. If a second event occurs at t < after the first event, it
will not be counted. However, in the paralysable case, if t < , the second event
will extend the dead time with a period from the moment of its interaction. If a
third event occurs at t > after the first event but within a period of time after
the second event, it will not be counted either. It will add another period of . The
dead time is not of fixed length. It can become much larger than the basic period
and in this case it is referred to as extendable dead time. Only if an event
occurs at time > after the previous event will it be counted. In this case, the
counting rate is the rate of occurrences of time intervals > between events, for
which the following can be derived:
R =Te T (6.3)
Figure6.2 demonstrates the relation between R and T for the two cases
above and for the case of =0, i.e. R=T.
6.2. GAS FILLED DETECTORS
6.2.1. Basic principles
The mode of operation of a gas filled detector depends strongly on the
applied voltage. In Fig.6.3(a), the signal amplitude is shown as a function of the
voltage V. If upon interaction with radiation an energetic electron ploughs through
the gas, the secondary electrons produced will tend to drift to the anode and the
ions to the cathode (seeFig.6.1(a)). If the voltage is relatively low, the electric
200
1
Non-paralysable
1
1
Paralysable
1
FIG.6.2. Counting rate R as a function of true event rate T in the absence of dead time
(R=T), in the non-paralysable case and in the paralysable case.
field E is too weak to efficiently separate the negative and positive charges. A
number of them will recombine. The full signal is not observed this is in the
recombination region. Increasing the voltage, more and more electrons and ions
escape from recombination. The region of full ionization is now reached. For
heavier charged particles and at higher rates, this will happen at a higher voltage.
The signal will become constant over a wide voltage range. Typical operating
voltages of an ionization chamber are in the range of 5001000 V.
For the discussion of operation at stronger electric fields, cylindrical
detector geometry with a thin anode wire in the centre and a metal cylinder as
cathode (seeFig.6.3(b)) is introduced. The electric field E(r) is proportional to the
applied voltageV and inversely proportional to the radius r. At a certain voltage
VT, the threshold voltage, the electric field near the anode wire is so strong that a
drifting electron will gain enough energy to ionize a gas atom in a collision. The
proportional region is entered. If the voltage is further increased, the ionization
zone will expand and an avalanche and significant gas amplification are obtained.
At normal temperature and pressure, the threshold electric field ET 106 V/m. For
parallel plate geometry with a depth of ~1cm, this would imply that VT 10kV,
which is not practicable. Due to the r1 dependence, in the cylindrical geometry,
manageable voltages can be applied for proportional operation (13kV). As long
as the gas gain M is not too high (M 104), it is independent of the deposited
energy. This is referred to as the proportional region and proportional counter.
If the voltage is further increased, space charge effects will start to reduce the
effective electric field and, consequently, affect the gain. This process will start
at a lower voltage for the higher primary ionization density events. The limited
proportionality region is entered. With further increasing voltage, the pulse
201
CHAPTER 6
height will eventually become independent of the deposited energy. This is the
GeigerMller region.
Limited
proportional
region
Pulse height (log scale)
(a)
(b)
Cathode
Recombination
Geiger
Mller
region
Anode
wire
2E
E
Radiation
Full
ionization
Proportional
region
VT
Cylinder geometry
High voltage
FIG.6.3. (a) Pulse height as a function of applied high voltage for gas filled detectors;
(b)cylindrical detector geometry.
Instead of one wire in a cylindrical geometry, many equidistant parallel

anode wires at a pitch of 12mm can be positioned in a plane inside a box with
the walls as cathode planes. This multiwire proportional chamber (MWPC) is
employed in autoradiography. The technique of photo-lithography made it
possible to introduce micro-patterned detectors that operate analogously to
the MWPC. Examples are the micro-strip gas chamber and the gas electron
multiplier. Spatial resolutions are of the order of 0.1mm.
6.3. SEMICONDUCTOR DETECTORS
As shown in Fig.6.1(b), a semiconductor detector is a capacitor. If upon
interaction with radiation, electrons are lifted from the valence band into the
conduction band, the transport of the charge carriers in an applied electric field is
observed. However, if a voltage difference is supplied to electrodes on opposite
sides of a slab of semiconductor material, in general, too high a current will flow
for practical use as a detector. At room temperature, electrons are lifted from the
valence band into the conduction band by thermal excitation due to the small gap
(Egap 1 eV). The resulting free electrons and holes cause the current. A solution
is found in making a diode of the semiconductor, operated in reverse bias. Silicon
202
is used as an example. The diode structure is realized by means of semiconductorelectronics technology. Silicon doped with electron-donor impurities, called
n-type silicon, can be used to reduce the number of holes. Electrons are the
majority charge carriers. Silicon with electron-acceptor impurities is called p-type
silicon; the number of free electrons is strongly reduced. The majority charge
carriers are the holes. When a piece of n-type material is brought into contact
with a piece of p-type material, a junction diode is formed. At the junction,
a space charge zone results, called a depletion region, due to diffusion of the
majority charge carriers. When a positive voltage is applied on the n-type silicon
side with respect to the p-type side, the diode is reverse-biased and the thickness
of the depletion layer is increased. If the voltage is high enough, the silicon will
be fully depleted. There are no free charge carriers left and there is virtually no
current flowing. Only a small current will remain, the leakage or dark current.
To make a diode, n-type silicon is the starting material and a narrow zone
is doped with impurities to make a p+n junction, as indicated at the bottom of
Fig.6.1(b). The notation p+ refers to a high doping concentration. For further
reduction of the leakage current, high purity silicon and a blocking contact are
used, i.e. an n+ doping at the n-type side, also indicated in Fig.6.1(b). If the
leakage current is still problematic, the temperature can be decreased. The use
of high purity semiconductor material is not only important for reducing the
leakage current. Energy levels in the gap may trap charge carriers resulting from
the interaction with radiation and the energy resolution of a detector would be
reduced.
The above described approach is not the only way to make a detector.
It is possible to start with p-type material and make an n+p junction diode.
Furthermore, it is possible to apply a combination of surface oxidation and
deposition of a thin metal layer. Such contacts are called surface barrier contacts.
If the thickness of a detector is <1mm, it is even possible to use intrinsic
silicon, symbol i, with p+ and n+ blocking contacts on opposite sides (pin
configuration). For thicker silicon detectors, yet another method is used. In
slightly p-type intrinsic silicon, impurities are compensated for by introducing
interstitial Li ions that act as electron donors. The Li ions can be drifted over
distances of ~10mm. Furthermore, if the bandgap of a semiconductor is large
enough, just metal contacts will suffice.
Important parameters are the mobilities, e and h, and the lifetimes, e and
h, of electrons and holes, respectively. The drift velocity e,h in an electric field E
is given by the product of the mobility and the field strength. Consequently, for
a given detector size and electric field, the mobilities provide the drift times of
the charge carriers and the signal formation times. From the mobilities and the
lifetimes, information on the probability that the charge carriers will arrive at the
203
CHAPTER 6
collecting electrodes is obtained. The path length a charge carrier can travel in its
lifetime is given by:
e,h e,h = e,h e,h E (6.4)
If this is not significantly longer than the detector depth, charge carriers
will be lost.
6.3.2. Semiconductor detectors
Some properties of semiconductor detector materials of relevance for
nuclear medicine, viz. the density , effective atomic number for photoelectric
effect Zeff, Egap and W value, the mobilities e,h and the products of the mobilities,
and the lifetimes of the charge carriers, are presented in Table6.2.
Silicon is primarily of interest for (position sensitive) detection of
low energy X rays, particles and light quanta. The latter are discussed in
Section 6.4.2.2.
TABLE6.2. PROPERTIES
MATERIALS
Si (300 K)
(g/cm3)
Zeff
2.3
14
Si (77 K)
OF
SEMICONDUCTOR
Egap
(eV)
Wa
(eV)
1.12
Mobility
(cm2/Vs)
DETECTOR
Mobilitylifetime
(cm2/V)
ee
hh
3.6
1350
480
>1
~1
1.16
3.8
21000
11000
>1
>1
Ge (77 K)
5.3
32
0.72
3.0
36000
42000
>1
>1
CdTe (300 K)
6.2
50
1.44
4.7
1100
80
3103
2104
Cd0.8Zn0.2Te
(CZT-300 K)
~6
50
1.52.2
~5
1350
120
4103
1104
HgI2 (300 K)
6.4
69
2.13
4.2
70
5103
3105
See Section 6.1.4.2.
For X ray detection in the range of ~300 eV to 60keV, planar circular

Li drifted pin detectors notated Si(Li) are commercially available with a
thickness up to 5mm. Diameters are in the range of 420mm. For typical field
strengths of ~1000 V/cm, the drift times to the electrodes are on the order of
tens of nanoseconds. Energy resolutions (FWHM) at 5.9keV are ~130220 eV if
204
operated at 77 K. Position sensitive silicon detectors with a large variety of pixel

structures are commercially available. Silicon detectors are also used in personal
dosimeters.
Germanium, with its higher density and atomic number, is the basic
material for high resolution ray spectroscopy. Detectors are made of high purity
material. Large volume detectors are made of cylindrical crystals with their
core removed (coaxial geometry). High purity n-type or p-type is used with the
corresponding junction contacts on the outside and the blocking contacts on the
inside. Germanium detectors are operated at 77 K. Cylindrical detectors up to
a diameter of ~10cm and a height of ~10cm are commercially available. Drift
times to the electrodes can be as large as ~100 ns. Typical energy resolutions are
~1keV at 122keV ray energy and ~2keV at 1332keV.
Cadmium telluride (CdTe) and cadmium zinc telluride (CZT) are of interest
because their atomic number is significantly higher than that of germanium,
and room temperature operation is possible due to the larger bandgap. High
purity n-type or p-type material is used. The energy resolution is worse than
that of Ge detectors, e.g. 2.5% FWHM at 662keV. This is primarily due to the
relatively short lifetime of the holes, resulting in incomplete charge collection.
Electronic correction techniques are used and/or detectors with special electrode
configurations (small pixels or grids) are made to observe the electron signal
only. Detector dimensions are up to approximately 25mm25mm10mm.
Detectors of 25mm25mm5mm with 16 pixels16 pixels are available.
These detectors are used, for example, for innovation of SPECT.
In principle, HgI2 (mercury iodide) is an attractive material for efficient
ray detection because of the large density and high atomic number. Owing to
the relatively large bandgap, room temperature operation is possible. However,
the mobilities are low and charge collection, in particular of the holes, is
poor. Consequently, application is limited to detector thicknesses 10 mm.
Field strengths of 2500 V/cm are applied and analogous to CdTe and CZT,
methods are used to observe the electron signal only. Detector areas are up to
~30mm30mm.
6.4. SCINTILLATION DETECTORS AND STORAGE PHOSPHORS
Scintillation of a material is the prompt emission of light upon interaction
with radiation. In nuclear medicine, inorganic ionic crystals are most important.
They combine high density and atomic number with a fast response and a high
light yield, and large crystals can be grown. These crystals form the backbone for
205
CHAPTER 6
X ray and ray detection. Another group is formed by organic scintillators, viz.
crystals, plastics and liquids, which have a low density and atomic number, and
are primarily of interest for counting of particles. In some inorganic scintillator
materials, metastable states (traps) are created that may live from milliseconds to
months. These materials are called storage phosphors. Scintillators and storage
phosphors are discussed later in this section. However, as light detection is of
paramount importance, light sensors are introduced first.
6.4.2. Light sensors
6.4.2.1. Photomultiplier tubes
The schematic of a scintillation detector is shown in Fig.6.4(a). A
scintillation crystal is coupled to a PMT. The inside of the entrance window of
the evacuated glass envelope is covered with a photocathode which converts
photons into electrons. The photocathode consists of a thin layer of alkali
materials with very low work functions, e.g. bialkali K2CsSb, multialkali
Na2KSb:Cs or a negative electron affinity (NEA) material such as GaAs:Cs,O.
The conversion efficiency of the photocathode , called quantum efficiency,
is strongly wavelength dependent (seeFig.6.5). At 400 nm, =2540%. The
emitted electrons are focused onto the first dynode by means of an electrode
structure. The applied voltage is in the range of 200500 V, and the collection
efficiency 95%. Typical dynode materials are BeOCu, Cs3Sb and GaP:Cs.
The latter is an NEA material. If an electron hits the dynode, electrons are
released by secondary emission. These electrons are focused onto the next
dynode and secondary electrons are emitted, etc. The number of dynodes n is in
the range of 812. The signal is obtained from the last electrode, the anode. At
an inter-dynode voltage of ~100 V, the multiplication factor per dynode 5. In
general, a higher multiplication factor is applied for the first dynode, e.g. 1 10,
to improve the single-electron pulse resolution, and consequently the signal to
noise ratio. Starting with N photons in the scintillator and assuming full light
collection on the photocathode, the number of electrons Nel at the anode is given
by:
N el = 1 n1 N (6.5)
Gains of 106107 are obtained. A negative high voltage (10002000 V) is

often used with the anode at ground potential and care must be taken of metal
parts near the cathode. Furthermore, the detector housing should never be opened
with the voltage on. Exposure to daylight would damage the photocathode
permanently.
206
(a)
Window
Window PMT
photocathode
(b)
Window PMT
Microchannel plates
Reflector
Light
quantum
Canning
Scintillator
Optical coupling
Focusing electrodes
Photocathode anode
FIG.6.4. (a) Schematic of a scintillation detector showing a scintillation crystal optically

coupled to a photomultiplier tube (PMT); (b)schematic of a microchannel plate-photomultiplier
tube.
PMTs are available with a large variety of specifications, including circular,

square or hexagonal photocathodes. Cathode diameters are in the range of ~10 to
~150mm. A ~50mm diameter PMT has a length of ~150mm including contact
pins. Pixelated multi-anode PMTs exist as well. To optimize time resolution,
special tubes are made with almost equal electron transit times to the anode,
independent of the cathode position where an electron is emitted. Although the
electron transit time is of the order of 30 ns, the transit time spread standard
deviation is not more than ~250 ps, and the signal rise time ~1.5 ns.
A PMT aimed at ultra-fast timing is the microchannel plate (MCP)-PMT.
For electron multiplication, it employs an MCP structure instead of a dynode
configuration (seeFig.6.4(b)). An MCP (thickness: ~1mm) consists of a large
number of closely packed hollow glass tubes (channel diameter: 550 m).
The inner surface of the tubes is covered with a secondary emission material,
viz. PbO. The glass surfaces on the front and back side are covered with metal
contacts. The MCP is placed in a vacuum, and a voltage of ~1000V is applied
between the contacts, positive on the back side. An electron that enters a glass
tube on the front side will hit the wall and secondary emission will occur. The
secondary electrons will be pulled to the back side by the electric field, hit the
channel wall and produce secondaries, etc. Eventually, they will leave the tube
at the back. Electron multiplication of ~104 can be obtained. In an MCP-PMT,
two MCPs are used at a close distance. The glass tubes are at an angle, thus
preventing ions from gaining too much energy. This structure of two MCPs is
called a chevron. At voltages of ~3000 V, stable gains of the order of 106 are
obtained. The advantage of the MCP-PMT is the short path length of the
electrons, resulting in transit times of a few nanoseconds and transit time spreads
of ~100 ps. MCP-PMTs are commercially available with circular (~10mm
diameter) and square photocathodes, the latter with multi-anode structures. The
sensitivities range from 115 nm (MgF2 window) to infrared.
207
Quantum efficiency (%)
CHAPTER 6
Silicon
photodiode
UV
sensitive
Photomultiplier
tube
Wavelength (nm)
FIG.6.5. Quantum efficiency as a function of scintillation wavelength for a blue sensitive

photomultiplier tube (full line), a photomultiplier tube with sensitivity extended into the
ultraviolet (dashed extension) and a silicon photodiode.
6.4.2.2. Silicon based photon sensors

Although PMTs are used on a large scale in nuclear medicine, the relatively
large size, high voltages, small quantum efficiency and sensitivity to magnetic
fields are a reason to prefer the use of silicon photodiodes in some applications.
These diodes are usually of the pin structure (PIN diodes). They have a
thickness of ~2mm including packaging, and are circular, rectangular or square,
up to ~30mm30mm. Bias voltages are <150 V. The quantum efficiency
of silicon diodes can be >80% at longer wavelengths (Fig.6.5). The large
capacitance of 20300 pF, and leakage current, ~110 nA, are a disadvantage,
resulting in a significant noise level that negatively affects energy resolution in
spectroscopy.
An avalanche photodiode (APD) is the semiconductor analogue to the
proportional counter. A high electric field is created in a small zone where a
drifting electron can gain enough energy to produce an electronhole (eh)
pair. An avalanche will result. The critical field for multiplication is ~107 V/m.
The higher the voltage, the higher is the gain. Depending on the type, voltages
are applied in the range of 501500 V. Gains are in the range of M < 200 to
~1000. The gain lifts the signal well above the noise as compared with the silicon
diode. At a certain gain, the advantage is optimal. At very high electric fields,
spontaneous charge multiplication will occur. The corresponding voltage is
208
called the break-down voltage Vbr. For gains of M 105 106, an APD can be
used at voltages >Vbr, where it operates in Geiger mode. The pulses are equal in
magnitude. Signal quenching techniques have to be used. Circular and square
APDs are available with areas in the sub-square millimetre to ~1cm2 range.
Various pixelated APDs are available, e.g. of 4 pixels8 pixels at a pitch of
~2.5mm and a fill factor 40%.
In a hybrid photomultiplier tube (HPMT), the photoelectrons are accelerated
in an electric field resulting from a voltage difference of ~10kV, applied between
the photocathode and a silicon diode which is placed inside the vacuum enclosure.
The diode is relatively small, thus reducing the capacitance and, consequently,
the noise level. As the production of 1 eh pair will cost 3.6 eV, ~3000 eh pairs
are produced in the diode per impinging electron. Consequently, the signals from
one or more photons can be observed well separated. Equipped with an APD, an
overall gain of ~105 is possible. HPMTs have been made with pixelated diodes.
Window diameters are up to ~70mm.
The silicon photomultiplier (SiPM) is an array of tiny APDs that operate
in Geiger mode. The dimensions are in the range of ~20 m 20 m to
100 m 100 m. Consequently, the number of APDs per square millimetre
can vary from 2500 to 100. The fill factor varies from <30% for the smallest
dimensions to ~80% for the largest. The signals of all of the APDs are summed.
With gains of M 105106, the signal from a single photon can be easily observed.
By setting a threshold above the one electron response, spontaneous Geiger
pulses can be eliminated. The time spread of SiPM signals is very small, <100 ps.
Excellent time resolutions have been reported. Arrays of 2 pixels2 pixels
and 4 pixels4 pixels of 3mm3mm, each at a pitch of 4mm, have been
commercially produced. A 16 pixel16 pixel array of 50mm50mm has
recently been introduced. Blue sensitive SiPMs have a photon detection
efficiency of ~25% at 400 nm, including a 60% fill factor.
6.4.3. Scintillator materials
6.4.3.1. Inorganic scintillators
In an inorganic scintillator, the bandgap has to be relatively large to avoid
thermal excitation and to allow scintillation photons to travel in the material
without absorption (Egap 4 eV). Consequently, inorganic scintillators are based
on ionic-crystal materials. Three steps for the production of scintillation photons
are considered (Fig.6.6): (i) interaction of radiation with the bulk material
and thermalization of the resulting electrons and holes on the energy scale,
electrons end up at the bottom of the conduction band and holes at the top of
the valence band; (ii) transport of these charge carriers to intrinsic or dopant
209
CHAPTER 6
luminescence centres; (iii) interaction with these centres, i.e. excitation, relaxation
and scintillation. Using this model, the number of photons Nph produced under
absorption of a ray with energy E is:
N ph =
E gap
SQ (6.6)
The first term on the right is the number of eh pairs at the bandgap edge.
Typically, 2.5. S and Q are the efficiencies of steps (ii) and (iii).
Thermalization
Conduction band
Q
Relaxation
Trap
Excitation
Radiationless
recombination
Valence band
Excitation
Scintillation
Luminescence
centre
FIG.6.6. Energy diagram showing the main process steps in an inorganic scintillator.
For the most relevant scintillators, the wavelength at emission

maximum max, light yield Nph, best reported energy resolution at 662keV R662
and the decay time of the scintillation pulse are presented in the last four
columns of Table6.3. In the first columns, some material properties are given,
namely density, effective atomic number for photoelectric effect Zeff, attenuation
length at 511keV, 1/511, and the percentage of interaction by the photoelectric
effect at 511keV. These scintillators are commercially available. If hygroscopic,
they are canned with reflective material (Fig.6.4). Only BaF2 and BGO have
an intrinsic luminescence centre. The other scintillators have Tl+ or Ce3+ ions as
dopant luminescence centre. The cerium doped scintillators show a relatively fast
response of the order of tens of nanoseconds due to the allowed 5d 4f dipole
transition of the Ce ion. The transitions of the Tl doped scintillators are forbidden
and are, consequently, much slower. In general, mixed or co-doped crystals have
advantages for crystal growing, response time, light yield or afterglow effects.
Large variation is observed for light yields. This is mainly due to S < 1, i.e. there
are traps of different kinds, resulting in loss of eh pairs by non-radiative
210
transitions. Using Eq.(6.6) and the proper values of Egap, only LaBr3:Ce appears
to have S Q 1.
TABLE6.3. SPECIFICATIONS OF SOME INORGANIC SCINTILLATORS
Scintillator
Zeff 1/511 Photoelectric max

Nph
R662
(g/cm3)
(mm) effect (%) (nm) (photons/MeV) (%)
(ns)
NaI:Tla
3.67
51
29
17
410
41000
6.5
CsI:Tl
4.51
54
23
21
540
64000
4.3 800, 104
BaF2
4.88
220
310
1500
10000
0.8
600
Bi3Ge4O12
(BGO)
7.1
75
10.4
40
480
8900
300
LaCl3:Cea
3.86
49.5
28
15
350
49000
3.3
25
LaBr3:Ce
5.07
46.9
22
13
380
67000
2.8
16
YAlO3:Ce
(YAP)
5.5
33.6
21
4.2
350
21000
4.4
25
Lu0.8Y0.2Al:Ce
(LuYAP)
8.3
65
11
30
365
11000
Gd2SiO5:Ce
(GSO)
6.7
59
14.1
25
440
12500
60
Lu2SiO5:Ce,Ca
(LSO)
7.4
66
11.4
32
420
~36000
3643
Lu1.8Y0.2SiO5:
Ce (LYSO)
7.1
420
30000
40
23
12
230
18
Hygroscopic.
6.4.3.2. Organic scintillators crystals, plastics and liquids

The scintillation mechanism of organic scintillators is based on molecular
transitions. These are hardly affected by the physical state of the material.
There are pure organic scintillator crystals such as anthracene, plastics such
as polystyrene, and liquids such as xylene. Furthermore, there are solutions
of organic scintillators in organic solid (plastic) and liquid solvents. Typical
combinations are p-terphenyl in polysterene (plastic) and p-terphenyl in toluene.
There are also systems with POPOP (para-phenylene-phenyloxazole) added for
wavelength shifting. In general, organic scintillators luminesce at ~420 nm, have
211
CHAPTER 6
a light yield of ~10 000 photons/MeV of absorbed ray energy and the decay
times are about 2 ns. The scintillators are usually specified by a commercial code.
6.4.3.3. Storage phosphors thermoluminescence and
optically stimulated luminescence
A storage phosphor is a material analogous to an inorganic scintillator.
The difference is that a significant part of the interaction energy is stored
in long-living traps. These are the memory bits of a storage phosphor. The
lifetime must be long enough for the application considered. Readout is done
either by thermal stimulation (heating) or by optical stimulation. An electron is
lifted from the trap into the conduction band and transported to a luminescence
centre. The intensity of the luminescence is recorded. These processes have been
coined thermoluminescence and optically or photon stimulated luminescence.
Storage phosphors have been used for dosimetry for more than fifty years
(thermoluminescence dosimeter). In particular, LiF:Mg,Ti (commercial name
TLD-100) is widely used. The sensitivity is in the range of ~50 Gy to ~1 Gy. A
newer and more sensitive material is LiF:Mg,Cu,P (GR-200), with a sensitivity in
the 0.2 Gy to 1 Gy range. Recently, an optically stimulated luminescent material
has been introduced, Al2O3:C. The sensitivity is in the range of 0.3 Gy to 30 Gy.
Storage phosphors are also used in radiography.
Bibliography
INTERNATIONAL CONFERENCE ON INORGANIC SCINTILLATORS AND THEIR
APPLICATIONS, SCINT 2007, IEEE Trans. Nucl. Sci. 55 (2008) 10291564.
SCINT 2009, IEEE Trans. Nucl. Sci. 57 (2010) 11571520.
INTERNATIONAL WORKSHOP ON ROOM-TEMPERATURE SEMICONDUCTOR
X- AND GAMMA-RAY DETECTORS (15th workshop), IEEE Trans. Nucl. Sci. 54 (2007)
761880.
(16th workshop), IEEE Trans. Nucl. Sci. 56 (2009) 16971884.
KNOLL, G.F., Radiation Detection and Measurement, 4th edn, John Wiley & Sons, New York
(2010).
LEO, W.R., Techniques for Nuclear and Particle Physics Experiments, 2nd edn, Springer,
Berlin (1994).
PROCEEDINGS OF NUCLEAR SCIENCE SYMPOSIUM AND MEDICAL IMAGING
CONFERENCE (annually), IEEE Trans. Nucl. Sci. (recent volumes).
212
RODNYI, P.A., Physical Processes in Inorganic Scintillators, CRC Press, Boca Raton, FL
(1997).
SCHLESINGER, T.E., JAMES, R.B. (Eds), Semiconductors for Room Temperature Nuclear
Detector Applications, Academic Press, San Diego, CA (1995).
TAVERNIER, S., GEKTIN, A., GRINYOV, B., MOSES, W.M. (Eds), Radiation Detectors for
Medical Applications, Springer, Dordrecht, Netherlands (2006).
213
CHAPTER 7
ELECTRONICS RELATED TO NUCLEAR
MEDICINE IMAGING DEVICES
R.J. OTT
Joint Department of Physics,
Royal Marsden Hospital
and Institute of Cancer Research,
Surrey
R. STEPHENSON
Rutherford Appleton Laboratory,
Oxfordshire
United Kingdom
7.1. INTRODUCTION
Nuclear medicine imaging is generally based on the detection of X rays and
rays emitted by radionuclides injected into a patient. In the previous chapter,
the methods used to detect these photons were described, based most commonly
on a scintillation counter although there are imaging devices that use either gas
filled ionization detectors or semiconductors.
Whatever device is used, nuclear medicine images are produced from a
very limited number of photons, due mainly to the level of radioactivity that can
be safely injected into a patient. Hence, nuclear medicine images are usually
made from many orders of magnitude fewer photons than X ray computed
tomography (CT) images, for example. However, as the information produced
is essentially functional in nature compared to the anatomical detail of CT, the
apparently poorer image quality is overcome by the nature of the information
produced.
The low levels of photons detected in nuclear medicine means that
photon counting can be performed. Here each photon is detected and analysed
individually, which is especially valuable, for example, in enabling scattered
photons to be rejected. This is in contrast to X ray imaging where images are
produced by integrating the flux entering the detectors. Photon counting,
however, places a heavy burden on the electronics used for nuclear medicine
imaging in terms of electronic noise and stability.
214
ELECTRONICS RELATED TO NUCLEAR MEDICINE IMAGING DEVICES
This chapter will discuss how the signals produced in the primary photon
detection process can be converted into pulses providing spatial, energy and
timing information, and how this information is used to produce both qualitative
and quantitative images.
7.2. PRIMARY RADIATION DETECTION PROCESSES
As described in Chapter6, the methods used for the detection of X ray and
ray photons fall into three categories, namely the scintillation counter, gas filled
detectors and semiconductors. Each of these techniques provides several detector
types and requires different electronics to produce and utilize the signals.
7.2.1. Scintillation counters
Figure7.1 shows a block diagram of a scintillation counter using a phosphor
and photomultiplier combination, together with the basic electronics required
to produce analogue and digital signals used to create an image. Table6.3
shows that the phosphors used in nuclear medicine can produce 150067000
optical photons per megaelectronvolt of energy deposited in the crystal and the
light emission time can vary from less than 1 ns up to ~1 s. Additionally, the
amplification of the optical signal by a photomultiplier can vary by an order of
magnitude or more depending on the photocathode quantum efficiency and the
number of dynodes. From this, it can be seen that the pulses produced by the
scintillation counter can vary substantially in both shape and amplitude, and that
the electronic devices used to manipulate these signals must be flexible enough
to account for these variations.
Electronic
base and
preamplifier
analogue to
digital
converter
FIG.7.1. Block diagram of a scintillation counter and associated electronics.
If the signal from the photomultiplier tube (PMT) anode is small, a

preamplifier is needed prior to full amplification. This form of amplifier is usually
incorporated into the PMT electronic base tominimize the noise generated prior
to preamplification. Similar arguments apply to the use of solid state based light
sensors such as photodiodes when coupled to phosphors.
215
CHAPTER 7
Both PMTs and photodiodes require voltage supplies to produce signals

in the case of a PMT, this voltage supply can be 13kV as each successive
dynode typically requires 100200V to produce sufficient amplification of the
electron signal. For a photodiode, the voltage required to totally deplete the
device is usually a few tens of volts for a simple photodiode and more for an
avalanche photodiode (APD).
7.2.2. Gas filled detection systems
Gas filled imaging systems convert the energy deposited by a ray photon
directly into ion pairs. It takes 2535 eV to produce a single ion pair, so the
primary signal from 99mTc will be 40005000 electrons. This signal will be
amplified in the gas detector using a high voltage (a few kilovolts) to produce
an electron avalanche of typically 106107 in a multiwire proportional chamber
(MWPC) (Fig.7.2). Typical dimensions of these devices for medical imaging are
between 30 and 100cm laterally by 1020cm in the direction of the ray.
These signals will clearly also need amplification if they are to be used as
analogue and digital output pulses for image formation.
Amplifier
Incoming ray
Cathode wire plane
Ionization
Gas filled chamber

Anode wire plane
Amplifier
FIG.7.2. Schematic of a two-plane multiwire proportional chamber detecting a ray.
7.2.3. Semiconductor detectors

A simple diode can be thought of as a solid state ionization chamber where
the region between the pn junction acts as a reservoir of electronhole (eh)
pairs. Incoming radiation produces eh pairs in the diode, the number of which
is proportional to the energy deposited in the diode, and an array of diodes can
function as a radiation imaging device. The energy needed to produce an eh pair
is ~35 eV (seeTable6.2) and so the size of the pulses produced varies less than
216
for a scintillation counter. However, the signals will still require some form of
amplification to produce useful analogue or digital information.
7.3. IMAGING DETECTORS
Having briefly discussed the production of signals by the three major
ionizing radiation detection processes, it is necessary to understand how these
methods are used to produce images in nuclear medicine. The two main imaging
devices used are the gamma camera and the positron camera. For completeness,
autoradiography imaging of tissue samples containing radiotracers is also
described.
Generally, both gamma camera and positron camera systems use
scintillation counters as the primary radiation detector because the stopping
power for X rays and rays is good in the high density scintillating crystals
used. However, there have been some examples of cameras using MWPCs and
semiconductors, and a brief description is provided here.
7.3.1. The gamma camera
Invented by Hal Anger, the gamma camera is usually based on the use of
a single large area (e.g. 50cm40cm of NaI(Tl)) phosphor coupled to up to a
hundred PMTs. The camera (Fig.7.3) can detect rays emitted by a radiotracer
distributed in the body. The lead collimator placed in front of the scintillation
counter selects the direction of the rays entering the device and allows an image
of the biodistribution of the tracer to be made.
Signal processing electronics

Photomultipliers
Phosphor
Collimator
Incoming rays
FIG.7.3. Schematic of a gamma camera for detecting single photons.
217
CHAPTER 7
The PMTs produce a signal that is proportional to the light generated in

the crystal. Positional information can be obtained by comparing the size of the
signals from different PMTs, whereas the energy information is related to the sum
of the PMT signals. The accuracy of both the energy and the positional information
depends on the stability of the signal production and also on the electronics used.
The amplitude of analogue signals from the PMTs depends on both PMT and
electronic stability, and the noise generated in the signal amplification process. The
noise is kept as low as possible by amplification close to each PMT. The amplified/
summed signal from the photomultiplier can be converted to a digital pulse train
using an analogue to digital converter (ADC), where the number of pulses is
proportional to the pulse height or charge in the signal. This signal is used to select
events in which most of the energy of the ray is detected by the camera, a useful
tool to reduce the effect on image production of rays scattered in the body prior
to detection. It is important at this stage to include only those signals from PMTs
that provide information above the intrinsic noise level of the electronics this is
done using some form of signal thresholding, such as a comparator.
Traditionally, the individual PMT pulses are digitized close to the PMTs
and these signals are analysed using capacitor or resistor circuits to determine
the positional information that is then sent to the computer system to form the
image. The centroid of the energy/pulse height information provides a position
that is closely related to the point at which the ray enters the crystal. Recent
improvements for calculating this position based on the digital outputs from the
PMTs uses nearest neighbour calculations based on a stored reference map of
positional information. These methods provide more accurate estimates of the
incident radiation entry point but are more demanding computationally.
Thus, the accuracy of the image production process is very much determined
by the initial signal sizes and the subsequent amplification and digitization.
7.3.2. The positron camera
The positron camera is used to simultaneously detect the two annihilation
photons produced by positron emitting tracers distributed in the body. The
detectors are usually made of many thousands of small scintillating crystals
coupled to up to a thousand PMTs. This means that there are many more amplifiers
which may have to function at higher count rates than those used in a gamma
camera. The detection of these two rays requires the addition of coincidence
circuits used to select the pulses from a single annihilation event. Figure7.4
illustrates the format of a positron camera based on multiple scintillating counters
in which the signals can be read out to form a 3D image.
The main difference in the electronics between the positron camera and
the gamma camera is the large number of PMT channels involved and the count
218
rates achieved in both cases, factors of 1020 are not unusual. In addition, the
pulses from a positron camera must be carefully shaped to allow accurate timing
information to be made in coincidence circuits to ensure that both annihilation
photons from a single annihilation event are detected. Time jitter in the pulses
will affect this process, allowing random photons from multiple nucleic decays
to be included in the data acquisition. In addition, the recent introduction of
so-called time of flight cameras requires very accurate (sub-nanosecond) timing
to be made between the two annihilation photons.
D1
Amplifier
Energy 1
ADC
Digital 1
Coincidence
circuit
ADC
D2
Amplifier
Trigger
Digital 2
Energy 2
FIG.7.4. Typical signal processing in a phosphor/photomultiplier tube based positron camera

pulses from two opposing detectors in the camera array are amplified, digitized, checked for
coincidence and then used to provide positional and energy information for the event detected.
7.3.3. Multiwire proportional chamber based X ray and ray imagers

As shown in Fig. 7.2, an X or photon generates an ionization signal in
the gas that is detected by the anode and cathode wire planes. The high voltages
across the wire planes cause electron avalanches close to the nearest wires and
these signals can be detected and amplified at either end of the wire.
In the PETRRA positron camera (Fig. 7.5), the initial ray detection is
performed using blocks of barium fluoride. The vacuum ultraviolet produced
in the crystal photo-ionizes the gas producing electrons that are subsequently
amplified in the gas by a series of wire planes. The MWPC positional
information is read out using delay lines coupled to the cathode wires. Signals
are induced in the delay lines and detected at either end using amplifiers that
produce signals with low time jitter. These signals are passed to constant fraction
discriminators (CFDs) to produce fast timing signals and then to time digitizers.
The time difference between the arrival of the signals at the two ends of the
219
CHAPTER 7
delay line is measured by the time to digital converters (TDCs) and provides
the positional information the accuracy of this information depends on the
intrinsic properties of the delay lines and the spread of the signal at the wire
planes, and in this system is ~4mm. Pulses produced after the gas amplification
region are used to provide the fast coincidence trigger to read the data into the
computer a timing resolution of ~23 ns is readily achievable.
Coinc
From second detector
Gate
Data enable
CFD
CFD
TDC
TDC
Pulse height
B Gamp
CAC
B
Gamp
G
C
A
Coinc
CFD
TDC
Barium fluoride crystal

Gas amplification wire planes
Gate wire plane
Cathode wire plane
Anode wire plane
Coincidence box
Constant fraction discriminator
Time digitizer
FIG.7.5. Schematic of the pulse production and readout system for the PETRRA positron
camera wire planes are shown as dotted lines.
A further process tominimize the recording of single ray events in this

system is the use of an electronic gate in which the transport of ionization
signals is controlled by an additional wire plane. This allows passage of the
ionization to the anode/cathode part of the chamber only if two MWPC detectors
have been triggered in fast coincidence. Anode signals can be used to measure
the pulse height produced by each detected ray. Overall, the electronics of this
camera has to manage count rates of several megahertz.
Similar detectors have been developed for imaging low energy rays,
animal imaging and tissue autoradiography.
7.3.4. Semiconductor imagers
There have been several attempts to make nuclear medicine imaging devices
using semiconductors as a ray camera. The need for a high Z material means
that presently only germanium and cadmium zinc telluride (CZT) have potential
as the primary ray detector. Germanium (in the form of GeLi) has been used as
a 2D strip detector where the signals from amplifiers at the end resistor chains
220
could be used to determine the position of any interaction in the sensor. However,
the modest stopping power of the material coupled with the need for a cryostat to
reduce the intrinsic noise of the detector made this design impractical.
More practical systems based on room temperature operation of CZT
have been developed by GE (the Alcyone system) and Spectrum Dynamics (the
DSPECT system). In the case of the latter system designed specifically for cardiac
imaging, ~1000 individual small CZT crystals are coupled to a tungsten collimator
providing an intrinsic spatial resolution of 3.54.2mm full width at half maximum
and a sensitivity of approximately eight times that of a scintillator based camera
most of the increases in sensitivity are due to the collimator design.
Silicon photodiodes have been used as an alternative to PMTs for both
gamma camera and positron camera designs. Here, APDs have been coupled to
phosphors and because of their small size, a truly digital camera design is possible.
In practice, due to the cost of APDs, only small systems have been developed. The
recent development of silicon photomultipliers promises further improvements in
nuclear medicine imaging.
7.3.5. The autoradiography imager
Autoradiography is based on the use of radioactive labels to determine the
microscopic distribution of pharmaceuticals in tissues excised from humans
or animals. A major use in humans is to detect areas of malignancy or tissue
malfunction. In animals, the method is used to track the uptake of drugs, for instance.
The pharmaceuticals are usually labelled with long lived radiotracers that have a
short range emission or low energy X ray or ray emission. Typical examples of
tracers used are 3H, 14C, 32P, 33P and 125I. Autoradiography imagers are required to
detect the emissions with high efficiency as the levels of uptake in tissue samples
are often very low. The gold standard for tissue radiography is film emulsion which
produces a high resolution (m) image of tissues, although these detectors have
low efficiency for detecting the radiation involved. Images can take days to weeks
to produce and this can be a severe limitation if diagnostic information is desired.
Digital autoradiography systems based on the use of thin phosphors, gas filled
detectors and silicon wafers can be 50100 times more efficient although the spatial
resolution is limited to typically a few tens of micrometres.
A phosphor based imager may use a very thin (50100 m) material such as
GADOX or CsI(Tl) coupled to a high resolution sensor, such as a microchannel
plate, a charge coupled device or a complementary metal oxide semiconductor
APD (Fig.7.6).
221
D2
CHAPTER 7
igure 7.4
igure 7.6
igure 7.7
igure 7.8
Energy 2
Amplifier
Beta particle
Segmented CsI(Tl)
CMOS APD
Read out
Read out
FIG.7.6. Beta particle detection in an autoradiography system based on a segmented CsI(Tl)

phosphor coupled to a complementary metal oxide semiconductor avalanche photodiode
(CMOS APD).
C
The limitations of these devices are mostly the pixel size of the sensor
and the noise in the sensor. Amplifiers with low noise are required and room
temperature operation desirable. Such a device can have a resolution of <50 m.
MWPC based autoradiography imagers have been built in which the

sample is placed in intimate contact with the chamber gas such a device will
Vout
Ihave a resolution of a few hundred micrometres.
More recently, direct detection
of particles and X rays has been performed using charge coupled devices and
+ complementary metal oxide semiconductor APDs. The advantage of such devices
is that the spatial resolution can be improved further (down to a few micrometres)
and 3H can be imaged.
7.4. SIGNAL AMPLIFICATION
C
As discussed
above, the primary signals from the radiation detectors are
generally small and need to be amplified without the injection of high levels
R
of noise into the signal readout system.
A preamplifier is needed prior to the
C
main amplification process if the signals
from the detector are very small, for
example, when aRPMT has insufficient dynodes to provide a large output pulse.
Preamplifiers are usually mounted immediately next to or as part of the output
stage of the detector tominimize the noise produced prior to full amplification.
The main amplifier can then be used to maximize and shape the signal (via
current and/or voltage gain) without over-amplifying noise.
7.4.1. Typical amplifier
The output current from a PMT is directly proportional to the amount
of light received from the phosphor. Although the PMT amplifies the electron
signal produced at the photocathode by a large factor, the current produced at the
anode is still very small. Amplifiers for PMTs are specially designed to transform
this current into voltage which can be directly input into an analogue to digital
222
D1
Amplifier
ADC
converter or a comparitor. In order to achieve the optimum signal to noise ratio,

the output current pulse is integrated in a capacitor, the resulting voltage Coincidence
forming
circuit of
the output signal. The capacitor is normally arranged in the feedback circuit
a wide bandwidth voltage amplifier chosen to have high input impedance and
ADC(tens of kiloherz
an extremely small input current. As data rates can be high
to megaherz), the operational frequency range of the amplifier must be able
to cope with these rates. Ideally, the PMT anode is connected directly to the
D2 a high Amplifier
charge amplifier input, with
value resistor providing a DC return path.
Capacitative coupling can be a problem at low frequencies where the signal may
be
degraded.
Figure
7.4 The charge amplifier integrates the current from the PMT, producing
an output voltage pulse. Figure7.7 illustrates a typical charge amplifier where the
output voltage Vout is given by:
1
Vout =
C
Beta particle
Q
I (t ) d t =
(7.1)
C
The configuration has negative feedback that increases the effective input
Read out
capacitance by a factor equal to the gain of the amplifier. This ensures that almost
all of the current flows into the amplifier even though the PMT and wiring can
Figure
7.6 capacitance. The feedback also reduces the output impedance, so
have
significant
that the amplifier acts as a voltage source.
Vout
I
+
FIG.7.7. A directly coupled charge amplifier producing a voltage output by integrating the
Figure 7.7
current produced in the photomultiplier tube.
The shape of the output pulse is important for the measurement of both
analogue and digital information,
C and is defined by the output stage of the
amplifier (Fig.7.8). The amplified signal is first passed through a CR (high pass)
R the low frequencies,
filter which improves the signal to noise ratio by attenuating
R
Figure 7.8
223
Segmen
CMOS
Read ou
Vout
I
+
CHAPTER 7
Figure 7.7
which contain a lot of noise and very little signal. The decay time of the pulse is
also shortened by this filter.
C
R
R
Figure 7.8
FIG7.8. A CRRC pulse shaping circuit.
Before the output of the amplifier, the pulse passes through an RC (low pass)
filter which improves the signal to noise ratio by attenuating high frequencies,
which contain excessive noise. The pulse rise time is lengthened by this filter.
The combined effect produces a unipolar output pulse and with suitably chosen
values, has an optimal signal to noise ratio.
7.4.2. Properties of amplifiers
The most important properties of an amplifier are gain, bandwidth, linearity,
dynamic range, slew rate, rise time, ringing, overshoot, stability and noise:
The gain of an amplifier is defined as the log ratio of the output power/
voltage Pout to the input power/voltage Pin and is usually measured in
decibels:
gain [dB]=10log(Pout/Pin) (7.2)
Gain in charge amplifiers is often expressed in millivolts output per
picocoulomb input charge.
The bandwidth of an amplifier is defined as the range of frequencies that
the amplifier operates and is often determined by frequencies at which the
power output drops to half its normal value (the 3 dB point). This is an
important feature of an amplifier attached to a high count rate detector as
required in positron emission tomography (PET) imaging, for example.
Amplifier linearity is limited when the gain of the amplifier is increased
to saturation point, resulting in output pulse distortion. Clearly, this is
important if the dynamic range of the pulses produced by the detector is
large. Dynamic range is defined as the ratio of the smallest and largest
224
useful output signals, with the former limited by the noise in the system and
the latter by amplifier distortion.
Rise time is often defined as the time taken for the output pulse to increase
from 1090% of its maximum and is a measure of the speed or frequency
response of the amplifier.
Slew rate is the maximum rate of change of the shape of the output
signal for the whole range of input signals, usually expressed in volts per
microsecond. This is very important if timing information is needed from
the detector as a poor slew rate will distort the bigger signals, making
them unsuitable for fast timing, as in PET. For PET applications, amplifier
rise times of the order of a few nanoseconds are needed, with no shape
distortion resulting from slew rate even on the biggest pulses.
Ringing is a problem when an amplifier produces a pulse that either
oscillates before reaching its maximum value or where the tail oscillates
before reaching the baseline. This can be a serious problem if timing
information is required or if the oscillations produce multiple triggers of the
output electronics downstream of the amplifier.
Stability is clearly an important parameter for an amplifier if the output
signals are to be used for either analogue or digital purposes. It is essential
that the amplifier output does not vary significantly for a given input
signal as the processes used to determine positional, energy and timing
information rely on the output for a given input being constant both in
offset, amplitude and shape. Factors that affect stability are numerous but
prime examples are variations in temperature, supply voltage and count rate
as well as long term drift.
Noise is a major impediment to the production of images using any of the
devices discussed above. Examples include thermal noise caused by the
thermal movement of charge carriers in resistors, shot noise caused by a
random variation in the number of charge carriers and flicker or 1/f noise
caused by the trapping or collisions of charge carriers in the structure of
the silicon used in the electronics. These sources combine to produce a
variation in the output signal of the combined detector/electronics system
that can affect the quality of images produced. The root mean square noise
of a system is defined as the square root of the absolute value of the sum of
the squares of the noise variances.
For a system using PMTs, the dominant noise component is that associated
with the number of photoelectrons produced at the photocathode as this is
amplified by the gain of the PMT dynode chain and subsequent electronics. For
a gas filled detector, the equivalent is the number of primary electrons produced
225
CHAPTER 7
at the first stage of the ionization process and for a silicon detector the important
parameter is the initial number of eh pairs produced.
7.5. SIGNAL PROCESSING
Once an amplified signal has been produced, it is then used to generate
both analogue and digital information about the detected event. The analogue
signal will relate to the energy deposited in the detector and is used, for example,
tominimize the number of scattered rays accepted into the image production
process. The digital signal is used to produce spatial and timing information.
7.5.1. Analogue signal utilization
The analogue information is generated by sending the pulse from the
amplifier into a single or multichannel pulse height analyser. In a gamma camera,
several energy windows are available, whereby the pulse height or charge is
compared with preset values that correspond to the known energies of the rays
being detected. In the simplest case for imaging a single energy ray emission,
two thresholds can be set to reject pulses that are above or below these values
(Fig.7.9).
Threshold 1
Threshold 2
Input pulse
Output pulse
FIG.7.9. A single channel pulse height analyser an output pulse is produced when the input
pulse is between the two thresholds. This system also functions as a single channel analogue
to digital converter.
When a radiotracer that emits several different energy rays is being used,
multiple thresholds can sort the information into several channels or images.
7.5.2. Signal digitization
Analogue signals are converted into digital signals that are subsequently
used to provide spatial and temporal information about each detected event.
226
Analogue
pulse
input
Ramp
generator
input
Enable
Counter
Clock
input
Digital pulse train to computer
FIG.7.10. Ramp-based single slope converter system for digitizing analogue pulses.
This is done using an ADC. The simplest method of digitizing an analogue

signal is by using a single slope converter (Fig.7.10). For this, a ramp signal is
generated and at the same time a clock producing digital output pulses is started.
When the ramp signal exceeds the input pulse, the clock is stopped and the
FIG.7.11. Pulse sequence from the system illustrated in Fig.7.10.
number of pulses generated corresponds to the amplitude of the signal. The faster
the clock, the higher the accuracy of the digitization achieved. This is a relatively
simple and low cost solution but is slow as the time taken to digitize the pulse is
2N clock cycles. The pulse sequence producing the digital output is shown
227
CHAPTER 7
in Fig.7.11. An important feature is that the analogue pulse shape must be

constant to allow accurate conversion. It is clearly possible to have more than one
ramp signal to provide several digitization regions if greater or lesser accuracy is
needed in any region.
A faster method of analogue to digital conversion is possible by using a
flash ADC. This is done using a large number of comparators (seeFig.7.12),
each having a different reference level. The output from each is the input into a
logic box that produces the multiple bits of the digital signal. If an N bit output
is needed, then 2N 1 comparators are needed. The method is fast as conversion
takes a single clock cycle but the system is complex and expensive and consumes
a lot of power. Typically, between 8 and 12 bits may be needed for nuclear
medicine imaging.
1st bit
2nd bit
Analogue
input
pulse
3rd bit
Nth bit
Reference level 2N 1
FIG.7.12. Schematic of a FLASH analogue to digital converter producing N bits of digital
data.
7.5.3. Production and use of timing information

In PET systems, the timing of events is very important as only pairs of
annihilation photons from the same radioactive decay contribute positively to
the image. As the single count rates in a PET scanner may be very high, fast
timing is required for coincidence imaging and time of flight measurement.
Coincidence timing systems can be based on the timing taken from the front edge
of two pulses, from a zero crossing point of the differentiated pulses or by using
a constant fraction method. The main problem with using a simple front edge
trigger is that the variation in pulse height of the analogue signals produces a
large variation in timing. Figure7.13 shows how the two pulses from detectors in
a PET system are used to generate a coincidence with CFDs.
228
Analogue
Input 1
Analogue
Input 2
FIG.7.13. The use of constant fraction discriminators (CFDs) to generate a fast coincidence
output the insert shows how the trigger point is set by a constant fraction of the pulse height.
The trigger points for the timing occur at a constant fraction of the shaped
analogue signal, so that the timing is not affected by the different signal pulse
heights. In this example, CFD1 generates a gate with a width set to more than
twice the measured timing resolution of the detectors. If the pulse from CFD2
falls within this gate, a coincidence (AND) output is generated; otherwise, the
event is rejected.
An alternative method of determining the timing from a pulse is to use the
zero crossing technique (Fig.7.14). In this method, the pulse is differentiated to
produce a bipolar pulse the timing is taken from the point where the pulse
crosses a reference line that is usually tied to ground hence, the zero crossing.
Again it is important that the pulse shapes are carefully controlled tominimize
jitter in the timing information.
FIG.7.14. A timing signal generated from the zero crossing point of a differentiated signal.
If the timing information is to be stored, then the two pulses from the
CFDs can be input into a TDC. In this case, the first pulse starts and the second
one stops a clock the number of pulses generated is proportional to the time
229
CHAPTER 7
difference between the pulses. If a fast clock is used, excellent timing information
is available for use in time of flight calculations, for example.
7.6. OTHER ELECTRONICS REQUIRED BY IMAGING SYSTEMS
7.6.1. Power supplies
Low voltage supplies are used to provide the power input for semiconductor
systems where a few tens of volts are sufficient. In some cases, batteries may
provide enough power but the need to maintain a constant current and voltage
makes this a modest solution. Usually, a low voltage supply converts mains AC
power, typically 240V (or 110 V), into DC voltages of, for example, 15V
and 5V to provide the line voltages for transistors and diodes. This is done by
combining a transformer, which reduces the voltage, and a rectifier, typically a
diode which allows only one half of the AC signal to pass this is half-wave
rectification (Fig.7.15).
Full-wave rectification is achieved by using a diode bridge that allows both
halves of the AC signal to be used, with one half being inverted. The oscillations
are removed using a filter, usually capacitors. The smoothest DC output is
provided by using a three phase AC input. The output is usually passed through
a voltage regulator to stabilize the voltage and remove the last traces of ripple.
Half-wave rectification
3-phase full-wave rectification
AC input
Full-wave rectification
FIG.7.15. Conversion of AC into DC using a transformer and rectifier system.
PMTs and MWPCs require power supplies that can provide voltages up to
several kilovolts. For example, each pair of dynodes in a PMT usually has at least
230
100V between them and even more may be used between the photocathode and
the first dynode to maximize the early gain of the PMT. These power supplies
usually have an oscillator and step up transformer operating at high frequency
to provide the drive plus a voltage multiplier consisting of a stack of diodes and
capacitors.
7.6.2. Uninterruptible power supplies
This form of support is needed for an imaging system to overcome loss of
power during periods of mains supply interruption. In this case, the output power
comes from the storage battery via some form of inverter. While the mains power
is available, it charges the battery as well as providing power to the imaging
device. If the mains supply is interrupted, the battery continues to provide support
to ensure that the imaging system can continue to be used. The size of the battery
support system depends directly on how long backup is needed or how long, for
example, it takes the operator to save data and shut down the system. As in most
imaging environments, the mains is replaced by a generator supply. The period
of support is often short but usually several hours of supply is available from an
uninterruptible power supply.
7.6.3. Oscilloscopes
In order to optimize the use of pulse generating equipment, an oscilloscope
is essential. This type of device allows the pulses from the detectors to be
displayed at various stages of generation prior to their use in image production.
For example, the pulse sequences illustrated above can be displayed on an
oscilloscope and this allows the equipment to be adjusted to provide the optimum
analogue and digital pulse sequence, shape and size.
An oscilloscope allows the pulses to be displayed on a 2D display, usually
with the vertical axis representing voltage (pulse height) and the horizontal
axis time. In addition to the amplitude of the pulses, the oscilloscope display
can be used to analyse the frequency of the signals being studied and also to
detect any pulse distortion such as oscillation or saturation. In an advanced form,
the oscilloscope can function as a spectrum analyser over a wide range of pulse
frequencies.
The original oscilloscopes were based on a cathode ray tube to display
the pulses but more modern systems use liquid crystal displays connected to
ADCs and other signal processing electronics. To the user, the oscilloscope will
present as a box with a display screen, input connectors and various controls.
The input from equipment can be done either directly using connecting cables/
sockets or through probes, often into a high impedance (e.g. 1M) or, for high
231
CHAPTER 7
frequency signals, 50 . The trace on the oscilloscope screen is adjusted by

various controls. The timebase control can adjust the horizontal display between,
for example, 10 ns up to seconds and the pulse height control from millivolts to
volts. Other controls include a beam finder, spot brightness and focus, graticule
control (to provide a visual measurement grid), pulse polarity and trigger level
controls, horizontal and vertical extent and position controls, selection of trigger
source (particularly useful for pulse coincidence display) and sweep controls to
provide single, multiple and delayed sweeps, for example. An example of an
oscilloscope that can be used for examining pulses from imaging equipment is
shown in Fig.7.16.
FIG.7.16. The front panel of the Tektronix 465 oscilloscope.
More recently, it has been possible to install oscilloscope software onto a

computer to provide a low cost solution for pulse display.
7.7. SUMMARY
The information provided above gives a general overview of the equipment
and electronics used in nuclear medicine imaging. In some cases, manufacturers
have developed special multichannel electronics readout systems tailored to the
detectors. These systems include the individual electronics elements described
above in a compact design that increases speed and accuracy. Such systems are
usually specific to the device involved.
232
BIBLIOGRAPHY
HOROWITZ, P., HILL, W., The Art of Electronics, Cambridge University Press (1982).
INIEWSKI, K. (Ed.), Medical Imaging: Principles, Detectors, and Electronics, John Wiley and
Sons, Hoboken, NJ (2009).
TURCHETTA, R., Electronics signal processing for medical imaging, Phys. Med. Imaging
Appl. 240 (2007) 273276.
WEBB, S., The Physics of Medical Imaging, Hilger (1988).
233
CHAPTER 8
GENERIC PERFORMANCE MEASURES
M.E. DAUBE-WITHERSPOON
Department of Radiology,
University of Pennsylvania,
Philadelphia, Pennsylvania,
United States of America
8.1. INTRINSIC AND EXTRINSIC MEASURES
8.1.1. Generic nuclear medicine imagers
The generic nuclear medicine imager, whether a gamma camera, single
photon emission computed tomography (SPECT) system or positron emission
tomography (PET) scanner, comprises several main components: a detection
system, a form of collimation to select rays at specific angles, electronics and a
computing system to create the map of the radiotracer distribution. This section
discusses these components in more detail.
The first stage of a generic nuclear medicine imager is the detection of the
rays emitted by the radionuclide. In the case of PET, the radiation of interest are
the 511keV annihilation photons that result from the interaction of the positron
emitted by the radionuclide with an electron in the tissue. For general nuclear
medicine and SPECT, there is one or sometimes more than one ray of interest,
with energies in the range of <100 to >400keV.
The rays are detected when they interact and deposit energy in the
crystal(s) of the imaging system. There are two main types of detector: crystals
that give off light that can be converted to an electrical signal when the ray
interacts (scintillators) and semiconductors, crystals that generate an electrical
signal directly when the ray deposits energy in the crystal. Scintillation detectors
include NaI(Tl), bismuth germanate (BGO) and lutetium oxyorthosilicate (LSO);
semiconductor detectors used in nuclear medicine imagers include cadmium zinc
telluride (CZT). Radiation detectors are described in more detail in Chapter6.
When a ray interacts in a scintillation crystal, it deposits some or all
of its energy. This energy is re-emitted in the form of light with a wavelength
dependent on the crystal material but not on the energy of the ray. The more
energy deposited in the crystal, the greater the intensity of the light emitted.
Scintillation crystals are coupled to photomultiplier tubes (PMTs), which serve
234
to convert the scintillation light into an electrical signal. If scintillation light

strikes the photocathode of the PMT, electrons are emitted from the photocathode
by the photoelectron effect. The number of photoelectrons emitted depends on
the intensity of the scintillation light and, therefore, the energy deposited in the
crystal. The energy required to produce a single photoelectron is ~1000 eV, so
only a few hundred to a thousand electrons are produced for each ray that
interacts, well below the number needed to produce a measurable current. The
PMT contains approximately ten stages that serve to increase the number of
electrons by secondary emission of electrons from these dynodes. The signal at
the output of the PMT is a measurable current, the amplitude of which is still
proportional to the energy deposited in the crystal.
Semiconductor detectors operate differently: the ray still deposits some
or all of its energy in the crystal through photoelectric absorption or, more
likely, Compton scattering interactions. However, that energy is not re-emitted
as scintillation light; instead, it creates electronhole (eh) pairs that are then
collected by application of an electric field to create a measurable signal. The
energy required to create an eh pair is ~3 eV, so many more charge carriers are
created in semiconductor detectors than in scintillators (seeChapter6).
While the exact implementations vary from system to system, the
electronics of nuclear medicine imagers have several common functions: they
determine the location of interaction of the ray in the detector, calculate the
energy deposited in the crystal and ascertain whether that energy falls within a
prescribed range of desirable energies, and for PET systems, measure the times
that the two annihilation photons interacted and evaluate whether the difference
in those times falls within a desired timing window to have both come from
the same annihilation event (i.e. from the same positron decay). If an event is
determined to be valid, its position (and sometimes the energy and timing
information) is sent to the computer to be stored along with the information for
the many other valid events.
In order to create an image of the distribution of radiotracer, the measured
locations of interaction of the rays must be converted to a 2D or 3D map
through image reconstruction. For 2D planar imaging with a stationary gamma
camera, this can be as simple as displaying the number of events at each detector
position. For PET or SPECT imaging, where measurements are made for many
views around the subject, the data must be combined through a reconstruction
algorithm. These techniques range from analytical methods such as filtered
back projection to iterative algorithms where estimates of the distribution are
calculated and refined based on a model of the imaging system. Not all events
accepted are actually useful events with accurate position, energy and timing
information. To obtain quantitative images (i.e. images whose counts are directly
235
CHAPTER 8
related to the amount of activity at each location), corrections must be applied for
these unwanted events as part of the reconstruction process.
Performance measures aim to test one or more of the components, including
both hardware and software, of a nuclear medicine imager.
8.1.2. Intrinsic performance
There are two general classes of measurements of scanner performance:
intrinsic and extrinsic. Intrinsic measurements reflect the performance of a
sub-part of the imager under ideal conditions. For example, measurements
made on a gamma camera without a collimator will describe the best possible
performance of the detector without the degrading effects of a collimator,
although the collimator is essential for clinical imaging. For a PET scanner,
intrinsic performance is often determined for a pair of detectors, rather than
the entire system. Intrinsic measurements are useful because they reflect the
best possible performance and can help isolate the source of any performance
degradations observed clinically. However, these measures are typically
performed under non-clinical conditions and will not reflect the performance
of the nuclear medicine imager for patient studies. Intrinsic measures also tend
to be measurements of an isolated characteristic of the system, rather than its
impact on imaging studies. They reflect the limits of performance achievable by
the detection system and electronics without collimators or image reconstruction.
8.1.3. Extrinsic performance
Extrinsic, or system, performance measures are made on the complete
nuclear medicine imager under conditions that are more clinically realistic,
although even these measures may not show the full clinical performance of
the system. On a gamma camera, extrinsic measurements are made with the
collimator in place; for SPECT and PET systems, the performance is often
measured on the reconstructed image. The extrinsic performance of a system
gives an indication of how well all of the components of the imager work together
to yield the final image. As most extrinsic performance measurements attempt to
isolate a single aspect of imaging performance (e.g. spatial resolution, count rate
performance, sensitivity), the conditions of these measurements generally do not
match the conditions encountered in patient imaging studies. However, the results
of extrinsic performance measurements are generally good indicators of clinical
performance or may provide useful information about system optimization for
clinical studies.
236
8.2. ENERGY RESOLUTION

8.2.1. Energy spectrum
The amplitude of the signal from the detector depends on the energy
deposited in the crystal. If the number of measured events with a given
amplitude is plotted as a function of the amplitude (Fig.8.1), the result is an
energy spectrum. The shape of the energy spectrum depends on the radiotracer
and rays emitted through its decay and the characteristics of the detector
material, but all energy spectra have common features. There is one (or more
than one) peak, called the photopeak, where the ray deposited all of its
energy in the detector through one or more interactions. There is also a broad,
lower energy region that reflects incomplete deposition of the rays energy
in the detector and/or Compton scattering of the rays in the body with the
subsequent loss of energy before detection. Even in the absence of scattering
material (i.e. for a point source in air), the photopeak is not a sharp peak but
is blurred. This broadening, which depends on the properties of the detector,
is due to statistical fluctuations in the detection of photons and conversion of
the energy deposited in the crystal into an electrical signal. This effect is larger
for scintillation detectors than for semiconductors. With scintillation detectors,
there are several steps in the conversion process that are subject to statistical
fluctuations, including the conversion of the rays energy into scintillation
light, collection of the scintillation light and conversion into photoelectrons at
the PMTs photocathode, and multiplication of those photoelectrons at each
dynode in the PMT. For semiconductor detectors, statistical uncertainty is
introduced in the number of eh pairs created when the ray deposits its energy
and in the collection of these pairs.
The goal of nuclear medicine imaging is to map the distribution of
radiotracers, so only rays that do not interact in the tissue before reaching the
detectors are useful; any rays that scatter in the body first change their direction
and do not provide an accurate measurement of the original radionuclides
location. Unscattered photons are those rays with energies in the photopeak.
Nuclear medicine imagers accept events whose energies lie in a window
around the photopeak energy in order to reduce the contribution of lower energy,
scattered rays. For PET scanners, a typical energy window is 440650 keV
for LSO detectors; for gamma cameras based on NaI(Tl) detectors, it is 15% of
the photopeak energy (e.g. 129.5150.5keV for 140keV rays from 99mTc, and
6882keV for the characteristic X rays from 201Tl with a 20% window).
237
CHAPTER 8
Number of events
Photopeak
FIG.8.1. An example of an energy spectrum, defined as the number of measured events with
a given amplitude plotted as a function of the amplitude, where the amplitude depends directly
on the energy deposited in the crystal.
8.2.2. Intrinsic measurement energy resolution

The intrinsic ability of a detector to distinguish rays of different energies is
reflected in its energy resolution. The energy resolution of a detector is defined as
the full width of the photopeak at one half of its maximum amplitude, divided by
the energy of the photopeak, and is typically expressed as a percentage. A smaller
energy resolution value means that the detector is better able to distinguish
between two rays whose energies are close to each other. The energy resolution
depends on the energy of the ray approximately as ( + E)1/2/E and, therefore,
the energy of the ray source must be specified when quoting the energy
resolution of a system. The energy resolution worsens at lower energies because
fewer photoelectrons are detected (in scintillation detectors) or eh pairs are
created (for semiconductors), so the statistical fluctuations in the measured signal
are greater. In addition, the energy resolution of a complete imaging system is
typically worse than that of small individual detectors due to slight differences in
operating characteristics between detectors.
238
Only rays that have not scattered in the body will provide accurate
information about the radiotracer distribution. Accordingly, the energy window
is optimal if it includes as many photopeak events as possible, since they are
more likely not to have interacted with the tissue, and as few lower energy events
as possible, since they are more likely to be the result of one or more Compton
scatter interactions in the tissue. As the energy resolution worsens, however, it
is necessary to accept more low energy events because the photopeak includes
lower energy rays. For example, for detection of 511keV annihilation photons,
the lower energy threshold for BGO (1520% energy resolution) was typically
set to 350380keV, while that for LSO (12% energy resolution) is 440460keV
and for LaBr3 (67% energy resolution) the lower energy threshold can be set as
high as 480490keV without loss of unscattered rays.
8.2.3. Impact of energy resolution on extrinsic imager performance
The energy resolution is an intrinsic measure of detector performance; it
defines theminimum width of the energy window for a given radiotracer. The
energy window in turn affects the amount of scattered photons accepted. The ratio
of scattered events to total measured events, the scatter fraction, is an extrinsic
performance characteristic that is of concern, especially for quantitative imaging.
In PET systems, for example, the clinical image is assumed to be linearly related
to the activity uptake; because scatter adds a smoothly varying background to the
image, it degrades the quantitative accuracy of the image and adds to the image
noise, even when accurately estimated and subtracted.
There are two major types of scattered event, those where the initial ray
scattered in the body and those where the ray was not completely absorbed in
the detector but instead scattered, losing some but not all of its energy. In both
cases, the measured energy of the ray is lower than the energy of the original
photon because some energy is given up to the electron, and the measured
position may no longer be related to the original source of the ray because the
scattered photon does not travel along the same direction as the original ray. For
typical patient sizes, scattering in the body is much more significant than detector
scattering.
The scatter fraction is an extrinsic performance measure that describes the
sensitivity of a nuclear medicine imager to scattered events. The measurement
involves imaging a line source in a uniformly filled phantom of a specified size
at a low activity level, where scattered and unscattered events can be reasonably
well differentiated. As the amount of scatter depends on the size and distribution
of scattering material in the scanner, the measured scatter fraction cannot be used
to infer the amount or distribution of scatter in patient images. However, it is a
good indicator of the relative sensitivity of the system to scatter.
239
CHAPTER 8
The scatter fraction is directly related to the energy resolution of the

system in the sense that the energy resolution determines the energy window,
in particular the lower energy threshold. This determines the imagers ability to
exclude scattered events. However, good energy resolution does not lead to a low
scatter fraction unless the energy window used is made appropriately narrow; a
scanner with 7% energy resolution will accept approximately as much scatter
as one with 12% energy resolution if both systems have the same lower energy
threshold. For this reason, measurement of the scatter fraction is a more clinically
relevant parameter than the energy resolution.
8.3. SPATIAL RESOLUTION
8.3.1. Spatial resolution blurring
The spatial resolution of a nuclear medicine imager characterizes the
systems ability to resolve spatially separated sources of radioactivity. An
individual point source of activity does not appear at a single pixel in the image;
rather, it is blurred over several pixels, largely due to statistical fluctuations in the
detection of the rays. Sources whose measured activity distributions overlap
cannot be distinguished as distinct sources and instead appear as a single, broad,
low contrast source.
In addition to blurring small structures and edges, resolution losses also lead
to a decrease in the contrast measure in these structures and at boundaries of the
activity distribution. Activity in small structures is blurred into the background
and vice versa. Areas of increased or decreased uptake are less easily detected
because of this loss of contrast (the partial volume effect).
In imagers composed of many small crystals, the spatial resolution of
the system is limited by the size of the detector elements. In gamma cameras
with a single, large crystal coupled to an array of PMTs, the spatial sampling
of the crystal determines the best spatial resolution achievable. The smaller the
crystal element or the more finely sampled the detector, the better an event can be
localized and the better the spatial resolution will be.
For a given size and sampling, crystals of different materials will have
different spatial resolutions. This is because rays do not interact at the surface of
a crystal but penetrate the crystal before interacting. If a crystal has a low density
and low atomic number Z, rays will travel further before interacting, compared
with a high density, high Z material. The ability to stop rays is referred to as
the materials stopping power; detectors with higher stopping powers will have
more accurate spatial localization than those with low stopping power because
there is less inter-crystal scatter. The effect of stopping power becomes more
240
apparent when rays enter the crystal at an oblique angle to the face of the crystal
(e.g. near the radial edge of a system comprising a ring of detectors). In that case,
the rays can completely pass through the entrance crystal before interacting in
a neighbouring crystal. The ray is then mis-positioned as though it had entered
the neighbouring crystal or in some intermediate location, depending on the
relative amounts of energy deposited by the two interactions.
Spatial resolution is also affected by the energy of the photon and, for
scintillation detectors, the efficiency of collection of the scintillation light by the
PMTs. The energy of the ray that is deposited in the crystal determines the
amplitude of the measured signal, which in turn defines how accurately it can be
localized in the detector. The spatial resolution measured in a given crystal with
99m
Tc (140keV) is inferior compared to that which would be measured with a
511keV photon.
As will be discussed later, the spatial resolution can also depend on the
count rate or amount of activity in the scanner. As the count rate increases, there
is an increased chance that two events will be detected at the same time in nearby
locations in the detector. These events will pile up and appear as a single event
at an intermediate location with a summed energy. This can lead to a loss of
resolution with increasing activity.
8.3.2. General measures of spatial resolution
There are several ways to characterize the spatial resolution, whether of a
detector or of a complete system. The point spread function (PSF) and line spread
function (LSF) are the profiles of measured counts as a function of position
across the point/line source. Rather than showing the complete profiles, however,
it is more convenient to characterize them by simple measures. The full width at
half maximum (FWHM) and full width at tenth maximum (FWTM) are useful
to describe the widths of the profile although they do not give information about
any asymmetry in the response. The equivalent width was defined as a way to
combine the FWHM and FWTM into a single parameter and describe the shape
of the profile in a simple way; it is defined as the width of a box function with
a height equal to the maximum amplitude of the profile and an area equal to
the total number of counts in the profile above 1/20 of its maximum amplitude.
Reducing the PSF or LSF to a few parameters carries with it a loss of information
about the spatial response of the imager; for example, LSFs or PSFs can have
very different shapes and still have the same FWHM.
The modulation transfer function (MTF) is one way to more completely
characterize the ability of a system to reproduce spatial frequencies. The MTF
is calculated as the Fourier transform of the PSF and is a plot of the response of
a system to different spatial frequencies. High spatial frequencies correspond to
241
CHAPTER 8
fine detail and sharp edges, while low spatial frequencies correspond to coarse
detail. The better the response at high frequencies, the smaller the structures
that can be resolved. A flat response across all spatial frequencies means that
the system most accurately reproduces the object. As it is difficult to compare
imaging performance based on the MTF, however, the FWHM and FWTM are
used to characterize spatial resolution.
8.3.3. Intrinsic measurement spatial resolution
The intrinsic spatial resolution is a measure of the resolution at the detector
level (or detector pair level for PET) without any collimation. It defines the
best possible resolution of the system, since later steps in the imaging hardware
degrade the resolution from the detector resolution. On gamma cameras, the
intrinsic resolution is determined using a bar phantom with narrow slits of
activity across the detector. On PET systems, the intrinsic resolution is measured
as a source is moved between a pair of detectors operating in coincidence. The
FWHM and FWTM of profiles of detected counts as a function of position are
taken as measures of the intrinsic spatial resolution. In both cases, the intrinsic
spatial resolution sets a limit on the resolution but does not translate easily into
a clinically useful value because other components of the imager impact the
resolution in the image.
8.3.4. Extrinsic measurement spatial resolution
The spatial resolution of a nuclear medicine imager depends on many
factors other than just the detectors. The linear and angular sampling play a
significant role: to preserve the intrinsic resolution, the imager should be sampled
every 0.1FWHM. Under-sampling leads to small structures being missed in
the image. For single-photon imagers, a collimator is used to limit the direction
of rays incident on the detector. Collimators are designed for specific purposes
(e.g. sensitivity or resolution) and/or specific radionuclides. As the hole size and
spacing of a collimator will affect the spatial sampling, each collimator will lead
to different system spatial resolution.
The reconstruction processing performed to create tomographic images
in SPECT or PET also affects the image resolution. Reconstruction algorithms
are generally chosen to preserve as much fine detail and edge information as
possible, while keeping image noise sufficiently low so that it is not confused
with actual structure. The parameters of reconstruction can, therefore, change
with the imaging study and with the number of events measured.
The spatial resolution is not constant throughout the imaging field of view
(FOV). For PET systems, the resolution does not vary significantly with location
242
of the source between two detectors in a detector pair, but the systems radial
resolution often degrades as the source is moved radially outwards from the
centre of the scanner. For gamma cameras, the resolution degrades as the source
is moved away from the detector face. For this reason, system spatial resolution
measurements are performed with the source at different locations in the imaging
FOV.
Extrinsic measures of spatial resolution are made under more clinically
realistic conditions and include the effects of the collimator (for single photon
imaging) and reconstruction processing. The extrinsic spatial resolution is
typically measured with a small point or line source of activity of a sufficiently low
amount such that effects seen at high count rates (i.e. mis-positioning of events)
are negligible. Measurements of system spatial resolution can be performed
in air or with scattering material added. A stationary source is positioned at
specified locations throughout the nuclear medicine imagers FOV. The spatial
resolution is determined from the images, including any reconstruction or
processing steps, by drawing profiles through the source. No spatial smoothing
or other post-processing is performed. In addition, any resolution modelling or
resolution recovery techniques applied during clinical reconstruction are not
used in the measurement of extrinsic resolution. The extrinsic spatial resolution
is distinguished from the intrinsic resolution because it includes many effects
not seen with the intrinsic resolution: collimator blurring, linear and angular
sampling, reconstruction algorithm, spatial smoothing, and impact of electronics.
While the extrinsic resolution measurement reflects the resolution of the
complete imaging system, the spatial resolution achieved in patient images
is typically somewhat worse than the extrinsic spatial resolution. The spatial
sampling is finer than occurs clinically because the pixel size is typically smaller
than that used for patient studies in order to sample the PSF or LSF sufficiently.
For imagers that reconstruct the data, the reconstruction algorithm in the
performance measurement is often not the technique applied to clinical data;
an analytical algorithm such as filtered back projection is generally specified
for tomographic systems to standardize results between systems. Another key
determinant of the clinical resolution is noise in the data that necessitates noise
reduction through spatial averaging (smoothing), which blurs the image. For
data with high statistics, a sharp reconstruction algorithm can be applied, and the
resulting image has good spatial resolution. For more typical nuclear medicine
studies, where the number of detected events is limited, some form of spatial
smoothing is applied, with the resulting blurring of fine structures.
243
CHAPTER 8
8.4. TEMPORAL RESOLUTION

8.4.1. Intrinsic measurement temporal resolution
As the activity in the FOV increases, events arrive closer to each other in
time until the imager cannot distinguish individual events. The timing resolution,
or resolving time, is the time needed between successive interactions in the
detector for the two events to be recorded separately. The timing resolution is
largely limited by the decay time of the crystal. For scintillators, the decay time
can be as high as 250300 ns or as low as 2040 ns, depending on the detector
material. Typically, the scintillation light does not decay with a single time
constant but with a combination of fast (nanosecond) and slow (microsecond)
components. For semiconductor detectors, the decay time is much smaller. In
addition to the detector decay time, the various components of the electronics can
contribute to the loss of temporal resolution. The timing resolution is generally of
less interest than its impact on the count rate performance of the system.
The timing of events is critical for PET, where two annihilation photons
must be detected within a timing window to be recorded as a valid event. The
timing window must be set wide enough to measure valid coincidence events but
not so wide that many coincidences between uncorrelated annihilation photons
(random coincidences) are accepted. Coincidence timing electronics are
carefully designed so that a detectors signal is processed as quickly as possible,
rather than waiting for the entire scintillation light to be measured. This allows
the coincidence timing window to be set to <10 ns, limited by the time of flight
of the two annihilation photons across the imagers diameter. For a ring diameter
of 90cm, theminimum coincidence time window would be 6 ns.
Recent developments in PET technology allow for the difference in times
of arrival (time of flight) of the two annihilation photons to be measured. For
time of flight systems, the coincidence time window is still 46 ns but the time
of flight difference can be measured with a resolution of 300600 ps. This time
of flight information is used in reconstruction to localize the events. The timing
resolution is measured with a low-activity source of activity by recording a
histogram of the number of events as a function of time difference.
8.4.2. Dead time
The consequence of a finite timing resolution is a loss of counts measured
at higher activities. When two photons arrive within the resolving time of the
detector, the two photons are seen by the electronics as a single event. One or both
of the events may be lost, and the events are also mis-positioned in space. The
random nature of radioactive decay means that there is always a possibility that
244
two events will arrive within the resolving time of the detector; this possibility
increases as the activity in the imager increases.
There are two kinds of dead time: non-paralysable and paralysable (seealso
Chapter6). Non-paralysable dead time arises when an event causes the system
to be unresponsive for a period of time, so that any later events that arrive
during that time are not recorded. For paralysable dead time, the second event
is not only not recorded but also extends the period for which the electronics
are unresponsive. At moderate count rates, paralysable and non-paralysable dead
times are the same; it is only at high count rates that the two types of dead time
differ (seeFig.8.2). It can be seen that systems with non-paralysable dead time
saturate at high count rates, while those with paralysable dead time peak and then
record fewer events as the activity increases. This leads to an ambiguity in the
measured count rate: the same observed count rate corresponds to two different
activity levels. The system dead time performance of nuclear medicine scanners
is typically intermediate between paralysable and non-paralysable dead time
because some components have paralysable dead time while other components
have non-paralysable dead time.
Measured count rate
No dead time
Non-paralysable
dead time
Paralysable
dead time
True count rate

FIG.8.2. System dead time as a function of count rate.
245
CHAPTER 8
With increased dead time, additional activity injected in the patient does not
lead to a comparable improvement in image quality or reduction in image noise.
Dead time losses depend on the single event rate, coincidence count rate (for
PET), and the analogue and digital design characteristics of the nuclear medicine
imager. Dead time losses can depend on the activity distribution, especially for
PET because of the different single photon and coincidence rate relationship
with source distribution. They also depend on the radioisotope because dead
time results from all rays that interact in the detector, not just the photons that
fall within the energy window. For imaging studies with a large dynamic range
(e.g. cardiac scans), count rate performance is critical.
To correct for event losses due to dead time, a correction based on a
decaying source study is often applied to clinical data. The dead time correction
will generally correct for the loss of counts, so that the number of counts in the
image is independent of the count rate; it does not, however, compensate for the
higher image noise that arises because fewer events are actually measured.
8.4.3. Count rate performance measures
The generic measurement of count rate performance involves determining
the response of the nuclear medicine imager as a function of activity presented
to the system. Typically, this requires starting with a high amount of activity and
acquiring multiple images over time as the activity decays. The energy window is
set at low activity levels and is not changed at higher activities to accommodate
a shift in the photopeak due to pile-up effects. By comparing the observed events
with the counts that would be expected after decay correction of events detected
at low activities, the system dead time can be determined as a function of activity
level. It is especially important to determine the maximum measurable count
rate, since higher activities would result in no increase and perhaps a decrease
in detected counts. While most count rate performance measures call for starting
with a high activity and imaging as the activity decays, if too high an activity
is used at the beginning of the measurement, the detector may show effects of
saturation during later measurements at lower activities. Therefore, the amount
of activity at the beginning of the study must be sufficient to measure the peak
count rate but not be so high as to saturate the system for a significant period.
Intrinsic count rate performance measurements are performed with a source
in air and without any detector collimation. This is typically performed only on
gamma cameras. The system, or extrinsic, count rate performance is measured
with the complete system, including any collimation or detector motion, and a
distributed source with scattering material (e.g. a cylindrical phantom of specified
dimensions or a source placed within scattering material). The scatter adds low
246
energy photons that contribute to pile-up and dead time that are not present in the
intrinsic measurement.
For PET, random coincidences also increase as the activity increases;
whereas the true coincidence rate would increase linearly with activity in the
absence of dead time losses, the random coincidence rate increases quadratically
with activity, so that their impact becomes greater at higher count rates. The
activity where the random rate equals the true event rate is of importance, in
addition to the activity and count rate at which the true count rate saturates or
peaks. A global measure of the impact of random coincidences and scatter on
image quality is given in the noise equivalent count rate (NECR) defined as:
NECR =
T2
(8.1)
T + S + kR
where T, S and R are the true, scatter and random coincidence count rates,
respectively, and k is a factor that is equal to one if a smooth estimate of random
coincidences is used and two if a noisy estimate is used. This parameter does not
include reconstruction effects or local image noise differences but can be useful
in determining optimal activity ranges.
For systems that correct for dead time, it is important to apply dead time
correction and to reconstruct the data in addition to looking at the count rates.
The quantitative accuracy of the dead time correction is determined by looking at
a large region of interest in decay-corrected, reconstructed images; the counts in
the region of interest should be independent of activity level. It is also important
to examine the images at high activities for artefacts that may arise due to
spatially-varying mis-positioning effects or inaccuracies in various corrections
with increased activity.
8.5. SENSITIVITY
8.5.1. Image noise and sensitivity
Images from nuclear medicine devices are typically noisy because the
amount of activity that can be safely injected and/or the scan duration without
patient discomfort or physiological changes in activity distribution is limited. The
number of detected events for a given amount of activity in the imaging systems
FOV is an important performance characteristic because a more efficient imager
can achieve low image noise with lower injected activity than a less efficient
system. Noise in the image can affect both visual (qualitative) image quality
247
CHAPTER 8
and quantitative accuracy, especially in areas of low uptake or low contrast.

The relative response of a system to a given amount of activity is reflected in its
sensitivity.
The sensitivity of a system is determined by many factors. The geometry of
the imager, especially the solid angle of the detectors, as well as any collimation
will determine how many photons reach the detectors. The stopping power and
depth of the detectors will impact how many of these photons are detected. In
addition, the radionuclides energy, coupled with the imagers energy resolution
and energy window, affect the number of accepted events. Finally, the number
of counts measured in a given time for a fixed amount of activity depends on the
source distribution and its position in the imager.
8.5.2. Extrinsic measure sensitivity
All performance measurements of sensitivity are extrinsic; for single
photon imaging, in particular, the collimator is a major source of loss of events,
so it is more clinically interesting to know the sensitivity of the system with a
particular collimator.
As noted above, the number of observed counts depends greatly on
the activity distribution. For this reason, any measurement of sensitivity is
performed under prescribed conditions that do not attempt to replicate patient
activity distributions. The source configurations and definitions of sensitivity
vary widely, however. For planar imaging, a shallow dish source without
intervening scatter material is used, and the sensitivity is reported as a count rate
per activity. For SPECT, a cylindrical phantom is filled uniformly with a known
activity concentration, and the sensitivity is reported as a count rate per activity
concentration. For whole body PET scanners, a line source that extends through
the axial FOV is imaged with sequentially thicker sleeves of absorbing material,
and the data are extrapolated to the count rate one would measure without any
absorber; the sensitivity is then reported as a count rate per unit activity. Small
animal PET systems use a point source in air centred in the scanner, and the count
rate per activity, as well as the absolute sensitivity (in per cent) are reported.
None of these sensitivity measurements can be used to predict the number of
events that will be observed for patient studies; however, systems with higher
sensitivity will generally record more events from a patient activity distribution
than those with lower sensitivity.
248
8.6. IMAGE QUALITY

8.6.1. Image uniformity
The uniformity of response of a nuclear medicine imager across the FOV is
important for both qualitative and quantitative image quality. All PMTs of a given
type do not respond exactly the same way, and a correction for this difference in
gain is applied before the image is formed. Collimators can also have defects that
lead to non-uniformities in the image. For tomographic scanners, corrections for
attenuation and unwanted events such as scatter can also affect the uniformity of
the image.
Intrinsic uniformity is measured without a collimator by exposing the
detector to a uniform activity distribution (e.g. from a distant, uncollimated
point source). Intrinsic uniformity is measured at both low and high count rates,
where mis-positioning effects become more pronounced. The extrinsic system
uniformity is determined with a collimator in place (for single photon imaging),
and images are processed or reconstructed as for clinical studies. In both cases,
sufficient counts must be detected, so that image noise is low. Quantitative
assessment of image uniformity includes variation of pixel counts in small
regions across the FOV. However, because simple metrics of non-uniformity
such as this do not provide a complete assessment of what the eye perceives in
the image, a visual analysis is also important.
8.6.2. Resolution/noise trade-off
Most performance measurements are carried out under non-clinical
conditions to isolate an aspect of the imagers performance. To include more
of the effects seen in clinical data, some performance standards call for a
measurement of image quality. The activity distribution is a series of small
structures (e.g. spheres of varying diameters) in a background activity typical of
the activity levels seen in patient studies. The activity is imaged for a clinically
relevant time, so that the noise level in the data is comparable to that in typical
patient studies. The data are processed in the same manner as clinical data. The
resulting image, then, is a better representation of the resolution and noise seen
clinically. Data analysis consists of such measures as sphere to background
contrast recovery, noise in background areas and/or signal to noise ratio in the
spheres. While still a simplistic and non-clinical distribution, the measurement
gives a more relevant indication of clinical resolution/noise performance.
249
CHAPTER 8
8.7. OTHER PERFORMANCE MEASURES

There are many other performance measures that reflect a given aspect of
a nuclear medicine imager. For planar systems, the spatial linearity, or spatial
distortion of the measured position of photons compared to the actual position, is
important for good image quality. A number of nuclear medicine imaging systems
incorporate anatomical (e.g. computed tomography or magnetic resonance
imaging) imagers into the scanner, and the images from the different modalities
must be registered spatially. Another area where spatial registration is necessary
is in single photon systems where multiple energy windows are used, and the
images acquired in the different windows must be overlaid to form the image.
Quantitative linearity and calibration is an important measurement for systems
such as PET scanners that aim to relate pixel values to activity concentrations.
250
CHAPTER 9
PHYSICS IN THE RADIOPHARMACY
R.C. SMART
Department of Nuclear Medicine,
St. George Hospital,
Sydney, Australia
9.1. THE MODERN RADIONUCLIDE CALIBRATOR
9.1.1. Construction of dose calibrators
Throughout the world, the instrument that is used in nuclear medicine to
measure radioactivity is the calibrated re-entrant ionization chamber, commonly
known as a radionuclide calibrator or dose calibrator. Commercial systems
comprise a cylindrical well ionization chamber connected to a microprocessorcontrolled electrometer providing calibrated measurements for a range of common
radionuclides (Fig. 9.1). The chamber is usually constructed of aluminium filled
with argon under pressure (typically 12 MPa or 1020 atm). Dose calibrators
with reduced gas pressure are available for positron emission tomography (PET)
production facilities where very large activities may be measured.
FIG.9.1. A typical dose calibrator (e.g. CRC 25R).
251
CHAPTER 9
A well liner, made of low atomic number material (e.g. lucite (Perspex))
which can be removed for cleaning, prevents the ionization chamber from
becoming accidentally contaminated. A sample holder is provided into which
a vial or syringe can be placed to ensure that it is positioned optimally within
the chamber. The dose calibrator may include a printer to document the activity
measurements or an RS-232 serial communications port or USB port to interface
the calibrator to radiopharmacy computerized management systems.
The chamber is typically shielded by the manufacturer with 6 mm of
lead to ensure low background readings. Depending on the location of the dose
calibrator, the user may require additional shielding, either to reduce background
in the chamber or to protect the operator when measuring radionuclides of
high energy and activity. However, this will alter the calibration factors due to
backscattering of photons together with the emission of Pb K shell X rays arising
from interactions within the lead shielding. If additional shielding is used, the
dose calibrator should be recalibrated or correction factors determined to ensure
that the activity readings remain correct.
As examples of commercial systems, the specifications of two widely used
dose calibrators are given in Table 9.1.
TABLE9.1. SPECIFICATIONS
CALIBRATORS
Specification
OF
TWO
COMMERCIAL
Capintec CRC-25R
DOSE
Atomlab 200
Ionization chamber
dimensions
26 cm deep 6 cm diameter
26.7 cm deep 7 cm diameter
Measurement range
Autoranging from 0.001 MBq

to 250 GBq
Autoranging from 0.001 MBq

to 399.9 GBq
Nuclide selection
8 pre-set, 5 user-defined
(80 radionuclide calibrations in
memory)
10 pre-set, 3 user-defined
(94 radionuclide calibrations in
manual)
Display units
Bq or Ci
Bq or Ci
Electrometer accuracy
<2%
1%
Response time
Within 2 s
1 s for activities >75 MBq
Repeatability
1%
0.3%
252
9.1.2. Calibration of dose calibrators

A dose calibrator can be calibrated in terms of activity by comparison with
an appropriate activity standard that is directly traceable to a national primary
standard. National primary standards are maintained by the relevant national
metrology institute, such as the National Physical Laboratory (NPL) in the United
Kingdom, the National Institute of Standards and Technology in the United States
of America and the Australian Nuclear Science and Technology Organisation
(ANSTO). Using the activity standard, a calibration factor for the ionization
chamber can be determined for the specific radionuclide. The reciprocal of the
calibration factor represents the efficiency N of the ionization chamber for the
radionuclide N.
The nuclide efficiency N can be expressed as the sum of two components:
N =
p i (E i ) i (E i )
(9.1)
where
pi(Ei) is the emission probability per decay of photons of energy Ei;
and i(Ei) is the energy dependent photon efficiency of the ionization chamber.
Figure 9.2 illustrates a typical efficiency curve as a function of photon
energy. Thin-walled aluminium chambers show a strong peak in efficiency
at photon energies around 50 keV. This results from the rapid increase of the
probability of photoelectric interactions in the filling gas with decreasing energy
and the low energy cut-off with aluminium walls at about 20 keV.
Knowing the energy dependent photon efficiency curve for a specific
ionization chamber will enable the nuclide efficiency for any radionuclide to be
determined from the photon emission probability for each photon in its decay.
The 511 keV annihilation radiation will be measured when the activity of
positron emitting radionuclides is to be assayed. A single calibration factor for all
positron emitters cannot be used as the emission probability of the positrons must
be taken into account. The probability (branching ratio) of positron emission for
11
C is 100% and for 18F is 96.7%.
253
CHAPTER 9
FIG.9.2. Efficiency curve as a function of photon energy.
9.1.3. Uncertainty of activity measurements

The following sections describe the major sources of uncertainty in dose
calibrator measurements.
9.1.3.1. Calibration factor
For medical radionuclides, such as 99mTc and 131I, the uncertainty of
national standards is typically in the range of 13%. However, when the standard
is used to calibrate a medical dose calibrator, the uncertainty will be larger due
to the inherent limit on instrument repeatability. Furthermore, the calibration
factor will be for the particular vial size and thickness, and volume of solution,
used for the national standard. The calibration factor for a different container
(a syringe) and/or a different volume may vary from the established calibration
by a significant amount (see Section 9.1.3.6).
9.1.3.2. Electronics
Electrometers measure the current output from the ionization chamber
ranging from tens of femtoamperes up to microamperes a dynamic range
of 108, corresponding to activity levels from kilobecquerels to hundreds of
gigabecquerels. Modern dose calibrators automatically adjust the range while
older units required the operator to select the appropriate range. The potential
for different linearity characteristics for each range may result in discontinuities
when the range is changed. The effects of inherent inaccuracy, linearity and range
changing are illustrated in Fig. 9.3. The linearity of the dose calibrator must be
254
established over the full range of intended use when the unit is commissioned
and verified as part of the quality control programme (see Section 9.2.1.2).
FIG.9.3. Electrometer inaccuracies (courtesy of the National Physical Laboratory).
9.1.3.3. Statistical considerations

Repeated measurements on a single sample will not be identical because
of the random nature of radioactive decay (see Chapter 5). If the measurement
period remains constant, the precision of the measured activity will increase as
the activity increases. Conversely, the precision will deteriorate for low activity
sources. To compensate for this, many calibrators automatically adjust the
measurement period depending on the activity level. This may vary from less
than one second to tens of seconds for low activities (<1 MBq).
9.1.3.4. Ion recombination
As the activity of the source increases, the probability of recombination
of the positive ions with electrons increases. At high source activities, this can
become significant and lead to a reduction in the measured current. The effect of
recombination is illustrated in Fig. 9.4. For most modern calibrators, the effects
of recombination should be less than 1% when measuring 100 GBq of 99mTc.
9.1.3.5. Background radiation
When the source holder is empty, the dose calibrator will still record
a non-zero reading due to background radiation. This will comprise natural
background and background from sources within the radiopharmacy. It could
255
CHAPTER 9
also be due to contamination on either the source holder itself or the well liner.
Most dose calibrators provide a background subtraction feature. An accurate
measurement of the existing radiation level is made by the calibrator (usually
integrating over several minutes to improve precision) which is then automatically
subtracted from each subsequent reading. This may lead to erroneous results if
the background radiation has changed since it was measured due to the presence
of additional nearby sources or contamination. It is, therefore, essential to make
regular checks of the background radiation level.
FIG.9.4. Effects of recombination (courtesy of the National Physical Laboratory).
9.1.3.6. Source container and volume effects

Variations in the composition and thickness of the source container will
give rise to corresponding variations in the measured activity. These effects
will be most noticeable for low energy photon emitters and pure emitters.
Measurements made at NPL, United Kingdom (Table 9.2) have shown that
variations in glass wall thicknesses, which were within the range of the vial
manufacturing tolerances, could lead to errors of up to 7% for 125I.
When the activity is drawn into a syringe, the source geometry will be
different from that in a vial. Not only will the composition and thickness of the
syringe wall be different from that of the vial, but the distribution of the source
will also be different depending on the size of syringe used. This is clearly evident
in Fig. 9.5, showing measurements at NPL for 111In in three sizes of syringe (1, 2
and 5 mL) from two different manufacturers in comparison to those measured in
a laboratory standard P6 vial. Also illustrated in Fig. 9.5 is the effect of changing
the source volume without changing the activity. Self-absorption of the emitted
256
TABLE9.2. REDUCTION IN DOSE CALIBRATOR RESPONSE DUE TO

INCREASES IN GLASS WALL THICKNESS OF 0.08 AND 0.2 mm
Radionuclide
Reduction in response with increase in vial wall thickness of

0.08 mm
0.2 mm
125
3%
7%
123
0.6%
1.5%
111
In
0.2%
0.4%
0.1%
0.25%
131
mL Sabre
mL Sabre
mL Sabre
mL Gillette
mL Gillette
mL Sabre
FIG.9.5. The effects of geometry and sample size on dose calibrator readings, demonstrated
for 111In measured in varying syringes (reproduced from Ref. [9.1]).
radiation will change as the source volume changes. This will be particularly
important for radionuclides with low energy components such as 123I. For 99mTc,
the correction will usually be less than 1% but should be confirmed for a new
dose calibrator or when the supplier of the syringes changes.
9.1.3.7. Source position
The manufacturers source holder is designed to keep the source at the area
of maximum response on the vertical axis of the well. Variations in response
due to changes in vertical height or horizontal position of a few millimetres are
usually insignificant.
257
CHAPTER 9
9.1.3.8. Source adsorption

Certain radiopharmaceuticals have been observed to adsorb to the surface
of the container. For example, up to 30% of the activity of 201Tl has been
found to be adsorbed onto the glass of P6 vials. 99mTc-tetrofosmin has been
shown to adsorb onto the surface of syringes, such that some types of syringe
may retain as much as 19% of the activity. Of this, 6% adhered to the rubber
plunger with the remainder attached to the plastic syringe barrel. Up to 15% of
99m
Tc-macroaggregate of albumin (MAA) may adhere to the syringe, although
the amount on the rubber plunger is usually no more than 1%. The possibility of
activity adsorption should be considered whenever the facility uses syringes from
a different manufacturer.
9.1.4. Measuring pure emitters
The detection efficiency of ionization chambers for radiation is
low as most, if not all, of the particles are absorbed in the source solution
(self-absorption), in the walls of the container or in the walls of the ionization
chamber. The dose calibrator response from particles will be almost entirely
from bremsstrahlung radiation (see Section 1.1.7). In the energy region of interest
for ionization chamber measurements, the bremsstrahlung photon spectrum
is roughly the same shape as the particle energy distribution. The average
particle energy is, therefore, a good parameter with which to characterize the
ionization chamber response to the bremsstrahlung radiation.
Bremsstrahlung radiation flux is proportional to the square of the atomic
number of the absorbing material. Thus, in argon-filled ionization chambers,
significant activities are required in order to obtain a precise estimate of the
activity. However, as substantial activities of radionuclides are required to be used
therapeutically, reliable measurements are possible using pure emitters used
clinically such as 90Y, 89Sr and 32P. However, geometry factors (see Section 9.1.3.6)
will be even more important and the system must be calibrated for the specific
containers and volumes to be used clinically. Manufacturers are now producing
dose calibrators specifically for the measurement of emitters. These use a sodium
iodide detector instead of an ionization chamber, resulting in a significantly
increased detection efficiency; however, as the manufacturers state in their product
literature, measurements still require exacting attention to the sample container,
the sample volume and activity concentration to achieve accurate results.
Most commercially available ionization chambers are provided with
calibration factors for commonly used emitters, although these will usually
correspond to the activity within a vial rather than a syringe. The type of vial used
in the calibration is often unspecified, so the user should verify the calibration in
258
the vials normally used in the practice. Similarly, the calibration of the activity
within the size of syringe to be used clinically should be established. Published
results comparing the intrinsic efficiencies of dose calibrators from five different
manufacturers found that all systems had a good calibration for 32P, a reduction
in efficiency of approximately 1020% for 89Sr, and a wide divergence in
efficiency for 90Y. For this radionuclide, the results obtained using the calibration
factors supplied by the manufacturers ranged from 64 to 144% of the true value,
re-emphasizing the need for the calibration to be confirmed within the nuclear
medicine department.
Several emitters used for radionuclide therapy include a ray component.
These radionuclides include 131I (364 keV, 81.5% abundance) and 186Re (137 keV,
9.5% abundance). For these radionuclides, the ionization chamber efficiency
is primarily determined by the contribution and the manufacturers supplied
calibrations will usually be accurate to within 10%.
9.1.5. Problems arising from radionuclide contaminants
Unfortunately, it is often not possible for a solution of a radionuclide to
be totally free of other radionuclides. The proportion of the total radioactivity
that is present as a specific radionuclide is defined as the radionuclide purity.
National and international pharmacopoeia specify the radionuclidic purity of
a radiopharmaceutical. For example, the European Pharmacopoeia entry for
67
Ga-citrate injection requires that no more than 0.2% of the total radioactivity
be due to 66Ga. This requirement must be met at all times up to the expiry time of
the product. The US Pharmacopoeia is less stringent, specifying that not less than
99% of the total radioactivity be present as 67Ga at the time of calibration.
The presence of contaminants, even when less than 1% of the total activity,
can have a marked effect on the ionization chamber current and, thus, on the
measured activity. The British Pharmacopoeia specification for 201Tl-thallous
chloride requires that Not more than 2.0 percent of the total radioactivity
is due to thallium-202 and not less than 97.0 percent is due to thallium-201.
Thallium-202 has a half-life of 12.2 d and the predominant photon energy is
440 keV. Another possible contaminant is 200Tl which has a half-life of 1.09 d
and prominent energies at 368 keV and 1.2 MeV. Both of these radionuclide
contaminants will have a high efficiency in a dose calibrator. As the half-life of
202
Tl is significantly longer than that of 201Tl, the relative proportion of 202Tl to
201
Tl will increase over time. If the accuracy of a dose calibrator is to be checked
with a 201Tl source, the apparent accuracy could change depending on when the
measurements are taken relative to the stated calibration date. The presence of
these high energy contaminants will have an adverse effect on image quality due
to increased septal penetration and will also lead to an increased radiation dose to
259
CHAPTER 9
the patient. The effective dose, in millisieverts per megabecquerel, for 200Tl, 201Tl
and 202Tl is 0.238, 0.149 and 0.608, respectively. It should be noted that these
problems will be increased if the radiopharmaceutical is administered prior to the
nominal calibration date, as the proportion of 200Tl will be higher.
9.2. DOSE CALIBRATOR ACCEPTANCE TESTING
AND QUALITY CONTROL
9.2.1. Acceptance tests
Acceptance tests for dose calibrators should include measurements of the
accuracy, reproducibility, linearity and geometry response. These are required to
ensure that the unit meets the manufacturers specifications and to give baseline
figures for subsequent quality control.
9.2.1.1. Accuracy and reproducibility
The accuracy is determined by comparing activity measurements using
a traceable calibrated standard with the suppliers stated activity, corrected for
radioactive decay. The accuracy is expressed in per cent deviation from the actual
activity and should be measured for all radionuclides to be used routinely. It is
recommended that measurements of a long lived source, for example 137Cs, be
recorded at the time of initial testing for each radionuclide setting to be used
clinically for later quality control.
The reproducibility, or constancy, can be assessed by taking repeated
measurements of the same source. If the sample holder is removed from the
chamber between each measurement, the measured reproducibility will include
any errors associated with possible variations in source position.
9.2.1.2. Linearity
There are several approaches to the measurement of the linearity response of
a dose calibrator. Typically, a vial containing a high activity of 99mTc is measured
repeatedly over a period of at least 5 d. During this time, a 100 GBq source will
decay to 0.1 MBq. It is essential that the initial activity represents the highest
activity that is likely to be used in clinical practice, which will usually be the
first elution from a new Mo/Tc generator. A semi-log plot of the measurements,
corrected for background, should follow the expected decay of the radionuclide.
Any deviation from the expected line at high activities indicates saturation of
response of the ionization chamber. Accurate background measurements, at the
260
time of each assay, are essential as the background will become an increasing
component of the reading as the source decays. Deviations from linearity at low
activities are likely to be due to radionuclide impurities, such as 99Mo in vials
containing 99mTc.
Another approach that can be used to check the linearity requires a series of
radioactive sources that cover the range of activities to be measured. The sources
should all be prepared from the same stock solution and the dispensed volumes
measured accurately by weighing the vials pre- and post-dispensing. The volume
of liquid in each vial should be adjusted with a non-radioactive solution, so that
the volume is identical in each vial, to eliminate any geometry dependency in
the measurement. The measured activities are corrected for decay to the time of
measurement of the first vial and plotted against the dispensed volumes to assess
the calibrator linearity. The error in this method will be increased if there are
any small variations in the vial wall thickness as the same vial is not used for all
measurements.
Finally, linearity can be assessed by repeated measurements on a single
vial using a series of graded attenuators appropriate for a specified test source to
reduce the measured ionization current. These are typically a series of concentric
cylinders that fit over the vial. The attenuation through each cylinder must be
accurately known to use this method.
9.2.1.3. Geometry
The measured activity may vary with the position of the source within
the ionization chamber, with the composition of the vial or syringe, or with the
volume of liquid within the vial or syringe. Appropriate correction factors must be
established for the containers and radionuclides to be used clinically, especially
if radionuclides that have a substantial component of low energy photons, such
as 123I, are to be used. For each vial or syringe to be used clinically, a series of
measurements should be undertaken in which the activity remains constant, but
the volume is increased from 10 to 90% of the maximum volume by the addition
of water or saline. Corrected for decay, a plot of activity against volume should
be a straight horizontal line. Any deviations from this can be used to calculate the
appropriate correction factor.
Similarly, vial to syringe correction factors can be determined by measuring
the activity transferred from the vial to the syringe (original vial activity minus
residual activity) and comparing this to the activity measured in the syringe itself.
Geometry dependencies should not change over time; however, if
the practitioner changes the manufacturer of the syringes or obtains the
radiopharmaceuticals in a different vial size, a new set of calibration factors
should be determined.
261
CHAPTER 9
9.2.2. Quality control

9.2.2.1. Background check
As noted in Section 9.1.3.5, when the source holder is empty, the dose
calibrator will still record an activity due to background radiation. This
will come from natural background, from sources within the radiopharmacy
and/or from contamination present on the source holder or well liner. It is a
useful practice to keep a spare source holder and a spare well liner, so that if
contamination is detected the contaminated item can be removed from service to
be decontaminated, or left until the radioactivity has decayed.
At a minimum, the background should be determined each morning before
the dose calibrator is used and recorded. The background subtraction feature,
if available, can be used at that time to remove the measured background from
subsequent measurements. The technologist should also confirm the absence of
any additional background before all activity measurements during the day.
9.2.2.2. Check source reproducibility
A long lived check source should be used on a daily basis to confirm the
constancy of the response of the dose calibrator. Sealed radioactive sources
of 57Co and 137Cs, shaped to mimic a vial, are available commercially for this
purpose. The check source should be measured on all radionuclide settings that
are used clinically. Although the recorded activity of a 137Cs source on the 99mTc
setting will not be a correct measurement of its activity, a reading outside of that
expected from previous results may indicate a faulty dose calibrator or a change
in calibration factor, in this case of 99mTc.
9.3. STANDARDS APPLYING TO DOSE CALIBRATORS
The International Electrotechnical Commission (IEC) has published two
standards [9.2, 9.3] and a technical report [9.4] relating to dose calibrators. IEC
standards are often adopted by national standards organizations. Reference [9.3]
is for manufacturers to use to ensure that the equipment performance is specified
in a standardized way, while Ref. [9.4] is aimed at the users of dose calibrators.
There should also be national standards covering dose calibrators. The
American National Standards Institute publication ANSI N42.13-2004 [9.5] is
often referenced by US manufacturers. This specifies the minimum requirements
in terms of accuracy and reproducibility for dose calibrators:
262
The accuracy of the instruments, at activity levels above 3.7 MBq shall be
such that the measured activity of a standard source shall be within 10%
of the stated activity of that source [9.5];
The reproducibilityshall be such that all of the results in a series
of ten consecutive measurements on a source of greater than 100 Ci
(3.7 106 Bq) in the same geometry shall be within 5% of the average
measured activity for that source [9.5].
9.4. NATIONAL ACTIVITY INTERCOMPARISONS
National metrology institutes are responsible for the development and
maintenance of standards, including activity standards. These institutes, often
in collaboration with the relevant national professional body, have undertaken
national comparisons of the accuracy of the dose calibrators used in clinical
practice. Such comparisons have used, where possible, the clinical radionuclides
67
Ga, 123I, 131I, 99mTc and 201Tl, and have been carried out in Argentina, Australia,
Brazil, Cuba, the Czech Republic, Germany, India and the United Kingdom.
In some countries, such as Cuba and the Czech Republic, participation in the
comparison is mandatory, while in many other countries it is voluntary. The
surveys can also be used to measure the reproducibility of the calibrators.
As an example, Table 9.3 shows the results from a survey undertaken in
Australia in 2007.
TABLE9.3. SUMMARY OF THE RESULTS OF THE DOSE CALIBRATOR
SURVEY UNDERTAKEN IN AUSTRALIA IN 2007
Radionuclide
99m
Tc
131
67
201
Ga
Tl
No. of calibrators
167
164
116
162
Within 5% error
86%
80%
84%
73%
Within 10% error
98%
95%
97%
94%
Within 10%
reproducibility
100%
100%
100%
100%
These surveys also offer the opportunity for the calibration factor to be
adjusted if a dose calibrator is found to be operating with an error of >10%.
263
CHAPTER 9
9.5. DISPENSING RADIOPHARMACEUTICALS

FOR INDIVIDUAL PATIENTS
9.5.1. Adjusting the activity for differences in patient size and weight
Protocols used in nuclear medicine practices should specify the usual
activity of the radiopharmaceutical to be administered to a standard patient. In
most western countries, the standard patient is taken to be one whose weight is in
the range 7080 kg. However, many patients fall outside of this range. If a fixed
activity is used for all patients, this will lead to an unnecessarily high radiation
exposure to an underweight patient and may lead to images of unacceptable
quality or very long imaging times in obese patients.
There have been various approaches to determining the activity to be
administered. These are usually designed to provide a constant count density in
the image to maintain image quality or to provide a constant effective dose to the
patient. For example, it has been shown that for myocardial perfusion scans using
99m
Tc-tetrofosmin, the activity should be increased by 150% for a 110 kg patient
and by 200% for a 140 kg patient in order to maintain image quality without
increasing imaging time.
It has been shown, using the radiation dose tables provided in International
Commission on Radiological Protection (ICRP) publications 53, 80 and 106
[9.69.8], that the effective dose (mSv/MBq) can be expressed as a simple
power function of body weight. Scaling factors for the activity, to give a constant
effective dose can, therefore, be derived from the expression (W/70)a, where W
represents the weight of the person and the power factor a is specific for the
radiopharmaceutical. Again, using 99mTc-tetrofosmin as an example, a is found
to be 0.834. Although the dosimetry models are only available up to 70 kg, this
power function can be extrapolated to derive scaling factors for patients whose
weight exceeds 70 kg. Using this approach, the activity should be increased by
146% for a 110 kg patient and by 178% for a 140 kg patient. This approach is
useful, but should be used with caution. The extrapolated activity would lead to
comparable organ and tissue doses for a patient of large body build but not for
a patient of similar weight due to large body fat deposits as the biodistribution
of the radiopharmaceutical would not be the same in these two cases. Table 9.4
presents the a value for common radiopharmaceuticals.
9.5.2. Paediatric dosage charts
Children are approximately three times more radiosensitive than adults, so
determining the appropriate activity to be administered for paediatric procedures
is essential. In addition to the scaling factor to be applied to the adult activity, a
264
TABLE9.4. THE POWER FACTOR a RELATING BODY WEIGHT TO

A CONSTANT EFFECTIVE DOSE ACCORDING TO THE EXPRESSION
(W/70)a FOR 14 COMMON RADIOPHARMACEUTICALS
Radiopharmaceutical
a value
99m
0.706
99m
0.801
Tc-DMSA
Tc-DTPA
99m
0.520
99m
0.849
Tc-MAG3
Tc-HMPAO
99m
Tc-MAA
0.871
99m
0.763
Tc-phosphonates
99m
a value
Tc-IDA
0.840
Tc-tetrafosmin
0.834
99m
99m
Tc-red cells
0.859
Tc-white cells
0.869
99m
18
0.842
99m
Tc-sestamibi
Radiopharmaceutical
F-FDG
0.782
Ga-citrate
0.931
I or 131I iodide
1.11
67
123
minimum activity must be specified in order to ensure adequate image quality.

In the past, the scaling factors were assessed using weight alone or body surface
area obtained from both height and weight. These two methods can give rise to
quite different scaling factors. For example, the scaling factor for a 20 kg child
is 29% of the adult activity using weight alone, but 43% when based on body
surface area.
Recently, the European Association of Nuclear Medicine (EANM)
Dosimetry and Paediatric Committees have prepared a dosage card which
recognizes that a single scaling factor is not optimal for all radiopharmaceuticals.
They used the methodology presented in Section 9.5.1 and were able to establish
that radiopharmaceuticals could be grouped into three classes (renal, thyroid and
others), with different scaling factors for each class. A dosage card is available on
the EANM web site that gives the minimum recommended activity and a weight
dependent scaling factor for each radiopharmaceutical which was determined to
give weight independent effective doses. This dosage card is reproduced here as
Fig. 9.6. To assist in these calculations, an on-line dosage calculator is available
on the EANM web site1, in which the user specifies the childs weight and the
radiopharmaceutical, and the recommended activity is displayed.
http://www.eanm.org/publications/dosage_calculator.php?navId=285
265
CHAPTER 9
9.5.3. Diagnostic reference levels in nuclear medicine

The recommendations of the ICRP specifically exclude medical exposures
from its system of dose limits, as the patient is directly benefiting from the
radiation exposure. However, in Publication 73 (1996) [9.9], the ICRP introduced
the term diagnostic reference level (DRL) for patients. DRLs are investigation
levels and are based on an easily measured quantity, usually the entrance surface
dose in the case of diagnostic radiology, or the administered activity in the case
of nuclear medicine. DRLs are referred to by the IAEA as guidance levels in
Safety Reports Series No. 40 [9.10], published in 2005. This publication contains
a table of guidance levels reflecting the values used in the early 1990s, when
single photon emission computed tomography procedures were far less common.
A survey of the use of DRLs in eight European countries, published in 2007
[9.11], showed that their introduction in nuclear medicine varied considerably.
For example, France had set DRLs for 10 nuclear medicine procedures, Germany
for 17 procedures, Italy for 48 procedures, while the United Kingdom listed 96
procedures. In some countries, the DRLs were set at the activities for which
marketing approval had been given, while in other countries the DRLs were
determined for each procedure by calculating the 75th percentile of the spread of
data values collected from a survey of participating practices. The latter approach
has been widely used to set DRLs in radiology and has been used in other parts
of the world, such as Australia and New Zealand, to establish DRLs in nuclear
medicine.
266
Dosage Card
(Version 1.2.2014)
Multiple of Baseline Activity
Weight
Class
Class
Class
Weight
Class
Class
kg
kg
Class
C
32
3.77
7.29
14.00
1.12
1.14
1.33
34
3.88
7.72
15.00
1.47
1.71
2.00
36
4.00
8.00
16.00
1.71
2.14
3.00
38
4.18
8.43
17.00
10
1.94
2.71
3.67
40
4.29
8.86
18.00
12
2.18
3.14
4.67
42
4.41
9.14
19.00
14
2.35
3.57
5.67
44
4.53
9.57
20.00
16
2.53
4.00
6.33
46
4.65
10.00
21.00
18
2.71
4.43
7.33
48
4.77
10.29
22.00
20
2.88
4.86
8.33
50
4.88
10.71
23.00
22
3.06
5.29
9.33
52-54
5.00
11.29
24.67
24
3.18
5.71
10.00
56-58
5.24
12.00
26.67
26
3.35
6.14
11.00
60-62
5.47
12.71
28.67
28
3.47
6.43
12.00
64-66
5.65
13.43
31.00
30
3.65
6.86
13.00
68
5.77
14.00
32.33
A[MBq]Administered = BaselineActivity x Multiple

a) For a calculation of the administered activity, the baseline activity value has to be multiplied by the multiples given above for the recommended radiopharmaceutical class
(see reverse).
b) If the resulting activity is smaller than the minimum recommended activity, the minimum activity should be administered.
c) The national diagnostic reference levels should not be exceeded!
Examples:
a) 18F FDP-PET Brain,
50 kg:
activity to be administered [MBq] = 14.0 x10.71 [MBq]

150 MBq
b) 123ImIBG,
3 kg:
activity to be administered [MBq] = 28.0 x1 [MBq] = 28 MBq

< 37 MBq (Minimum Recommended Activity)
activity to be administered: 37 MBq
This card is based upon the publication by Jacobs F, Thierens H, Piepsz A, Bacher K, Van de
Wiele C, Ham H, Dierckx RA. Optimized tracer-dependent dosage cards to obtain weightindependent effective doses. Eur J Nucl Med Mol Imaging. 2005 May; 32(5):581-8.
This card summarizes the views of the Paediatric and Dosimetry Committees of the EANM
and reflects recommendations for which the EANM cannot be held responsible.
The dosage recommendations should be taken in context of good practice of nuclear
medicine and do not substitute for national and international legal or regulatory provisions.
Android App
iPhone App
EANM Executive Secretariat

Hollandstrasse 14/Mezzanine 1020 Vienna, Austria
Phone: +43-1-2128030, fax: +43-1-21280309
office@eanm.org - www.eanm.org - fb/officialEANM
FIG.9.6. European Association of Nuclear Medicine (EANM) paediatric dosage card

(courtesy of EANM).
267
CHAPTER 9
Recommended Amounts in MBq

Radiopharmaceutical
Class
Baseline Activity
(for calculation
purposes only)
Minimum
Recommended
Activity1
MBq
MBq
I (Thyroid)
0.6
I Amphetamine (Brain)
13.0
18
123
123
I HIPPURAN (Abnormal renal function)
5.3
10
I HIPPURAN (Normal renal function)
12.8
10
I mIBG
28.0
37
I mIBG
5.6
35
F FDG-PET torso
25.9
26
F FDG-PET brain
14.0
14
F Sodium fluoride
10.5
14
Ga Citrate
5.6
10
Tc ALBUMIN (Cardiac)
56.0
80
Tc COLLOID (Gastric Reflux)
2.8
10
Tc COLLOID (Liver/Spleen)
5.6
15
Tc COLLOID (Marrow)
21.0
20
Tc DMSA
6.8
18.5
Tc DTPA (Abnormal renal function)
14.0
20
Tc DTPA (Normal renal function)
34.0
20
Tc ECD (Brain perfusion)
32.0
110
Tc HMPAO (Brain)
51.8
100
Tc HMPAO (WBC)
35.0
40
Tc IDA (Biliary)
10.5
20
Tc MAA / Microspheres
5.6
10
Tc MAG3
11.9
15
Tc MDP
35.0
40
Tc Pertechnetate (Cystography)
1.4
20
Tc Pertechnetate (Ectopic Gastric Mucosa)
10.5
20
Tc Pertechnetate (Cardiac First Pass)
35.0
80
Tc Pertechnetate (Thyroid)
5.6
10
99m
Tc RBC (Blood Pool)
56.0
80
Tc SestaMIBI/Tetrofosmin
(Cancer seeking agent)
63.0
80
Tc SestaMIBI/Tetrofosmin2
(Cardiac rest scan 2-day protocol min)
42.0
80
(Cardiac rest scan 2-day protocol max)
63.0
80
(Cardiac stress scan 2-day protocol min)
42.0
80
(Cardiac stress scan 2-day protocol max)
63.0
80
(Cardiac rest scan 1-day protocol)
28.0
80
(Cardiac stress scan 1-day protocol)
84.0
80
2.8
20
70.0
100
123
123
123
131
18
18
18
67
99m
99m
99m
99m
99m
99m
99m
99m
99m
99m
99m
99m
99m
99m
99m
99m
99m
99m
99m
99m
99m
99m
99m
99m
99m
Tc Spleen (Denatured RBC)
99m
99
Tc TECHNEGAS (Lung ventilation)3
The minimum recommended activities are calculated for commonly used gamma cameras or positron
emission tomographs. Lower activities could be administered when using systems with higher counting
efficiency.
The minimum and maximum values correspond to the recommended administered activities in the
EANM/ESC procedural guidelines (Hesse B, Tagil K, Cuocolo A, et al). EANM/ESC procedural guidelines for
myocardial perfusion imaging in nuclear Cardiology. Eur J Nucl Med Mol Imaging. 2005 Jul;32(7):855-97.
This is the activity load needed to prepare the Technegas device. The amount of inhaled activity will be lower.
FIG.9.6. European Association of Nuclear Medicine (EANM) paediatric dosage card

(courtesy of EANM) (cont.).
268
9.6. RADIATION SAFETY IN THE RADIOPHARMACY

9.6.1. Surface contamination limits
Surface contamination with radioactivity could lead to contamination of a
radiation worker and/or external irradiation of the skin of the worker. Internal
contamination could arise from inhalation and/or ingestion of the radionuclide.
The surface contamination limits given in Table 9.5 were derived based on
a committed effective dose limit of 20 mSv/a and the models for inhalation
and ingestion given in ICRP publications 30, 60 and 61 [9.129.14]. For each
radionuclide, the most restrictive pathway (inhalation, ingestion or external
irradiation) was used.
TABLE9.5. DERIVED LIMITS FOR SURFACE CONTAMINATION
Surfaces in designated
areas, including
protective clothing
(Bq/cm2)
Interiors of glove boxes

and fume cupboards
(Bq/cm2)
Non-designated areas
including personal
clothing (Bq/cm2)
18
100
1000
32
100
1000
51
1000
10000
50
67
1000
10000
50
89
100
1000
90
Nuclide
F
P
Cr
Ga
Sr
100
1000
99m
Y
Tc
1000
10000
50
111
In
1000
10000
50
123
1000
10000
50
125
100
1000
131
100
1000
177
1000
10000
50
201
1000
10000
50
I
I
I
Lu
Tl
269
CHAPTER 9
9.6.2. Wipe tests and daily surveys

Surveys of the radiopharmacy must be undertaken to ensure that these
surface contamination limits are not exceeded and that the operator is not
unnecessarily exposed to external radiation. Exposure could result from sources
inadvertently left on a bench and from contamination on bench surfaces.
Aerosolized droplets from a syringe during dispensing may go unnoticed, so it is
essential that all staff are aware that the dispensing area may be contaminated and
always wear protective gloves when working in this area. All radiopharmaceutical
elution, preparation, assay and administration areas should be surveyed at the end
of each working day.
Surveys should initially be undertaken with a survey meter to ensure that
no unexpected exposed sources are present in the radiopharmacy. All surfaces
should then be checked for contamination using a contamination monitor with
a probe appropriate to the radionuclides used. The background radiation levels
in the radiopharmacy, particularly in the dispensing area, are often higher than
elsewhere in the nuclear medicine department, so quantifying any contamination
found using a probe is difficult. If a low energy emitter is being used, it will
prove difficult or impossible to detect with an external probe. In these situations,
a wipe test should be used. A minimum area of 100 cm2 should be wiped and
then the activity on the wipe can be assessed using a pancake probe, or more
accurately in a well counter. For low energy emitters, such as 3H or 14C, liquid
scintillation counting must be used. When quantifying the surface contamination,
it is generally assumed that a wipe test using a dry wipe will remove one tenth of
the contamination while a wet wipe will remove one fifth of the contamination.
9.6.3. Monitoring of staff finger doses during dispensing
Systematic studies of the dose to the hands of staff working in
radiopharmacies have shown that finger doses may approach or exceed the annual
dose limit of 500 mSv to the extremities. The most exposed parts of the hands
are likely to be the tips of the index and middle fingers, and the thumb of the
dominant hand, with exposure for the index finger being highest. The ICRP has
recommended that finger dose monitoring be undertaken for any person handling
more than 2 GBq/d and regular monitoring should be carried out if doses to the
most exposed part of the hand exceed 6 mSv/month.
Although the dose to the finger tip will be the highest, it is much more
practical to wear a ring monitor at the base of the finger. A thermoluminescent
dosimeter chip mounted in a plastic ring is usually the most convenient type of
monitor. Such monitors are often available in a variety of sizes. The ring should
fit tightly, so that it is not inadvertently removed when the gloves are taken off.
270
The ICRP recommends that the ring monitor be worn on the middle finger with
the element positioned on the palm side, and that a factor of three should be
applied to derive an estimate of the dose to the tip. If the dosimeter element is
worn facing towards the back of the hand, a factor of six should be applied.
The dose to the fingers is critically dependent on the dispensing technique
used and the skill of the operator. It is important that staff undertake extensive
training in the dispensing technique with non-radioactive solutions prior to
dispensing radiopharmaceuticals for the first time. This is particularly important
with PET radiopharmaceuticals as the specific dose rate constant is much higher
for positron emitters than for radionuclides used for single photon imaging.
9.7. PRODUCT CONTAINMENT ENCLOSURES
9.7.1. Fume cupboards
A fume cupboard is an enclosed workplace designed to prevent the spread
of fumes to the operator and other persons. The fumes can be in the form of
gases, vapours, aerosols or particulate matter. The fume cupboard is designed
to provide operator protection rather than protection for the product within the
cabinet. A fume cupboard would, therefore, not be suitable as an area for cell
labelling procedures as this requires that the blood remain sterile at all times.
Fume cupboards usually include a transparent safety screen which can be
adjusted either vertically (more commonly) or horizontally to vary the size of the
working aperture into the cabinet. Some cupboards are available with a lead glass
safety screen to minimize the need for additional radiation shielding. The most
common type of fume cupboard is known as a variable exhaust air volume fume
cupboard which maintains a constant velocity of air into the cabinet (the face
velocity) irrespective of the sash position. Figure 9.7 shows a fume cupboard
suitable for use with radioactive materials.
Fume cupboards are available which discharge the exhaust air directly,
or after carbon filtration, to the atmosphere, usually above the building. Other
cabinets, known as recirculating fume cabinets, rely on filtration or absorption
to remove airborne contaminants released in the cabinet, so that the air may be
safely discharged back into the laboratory. Recirculating fume cabinets are not
normally applicable for use with radioactive materials.
Any installed fume cupboards must meet the requirements of the local
appropriate standard and any air discharged to the atmosphere must meet the
requirements of the appropriate regulatory authority. The standard will usually
specify the minimum face velocity through the working aperture (e.g. 0.5 m/s).
This should be checked on a regular basis and should be measured with the
271
CHAPTER 9
aperture fully open and in its minimum position. At the minimum position, the
face velocity may need to be higher to retain a constant exhaust rate from the
cabinet.
Before initial use, and as part of a regular quality control schedule, a smoke
test should be performed. This is to provide visual evidence of fume containment
within, or escape from, the fume cupboard. Smoke is released in and around the
fume cupboard and the visual pattern of airflow is observed. The results of the
smoke test must be documented and any reverse flows from the confines of the
cupboard corrected before subsequent use.
FIG.9.7. Fume cupboard suitable for use with radioactive materials.
9.7.2. Laminar flow cabinets

Laminar flow cabinets provide a non-turbulent airstream of near constant
velocity, which has a substantially uniform flow cross-section and with a variation
in velocity of not more than 20%. Laminar flow cabinets provide product
protection while a fume cupboard is designed to provide operator protection. The
air supplied to the cabinet is usually passed through a high efficiency particulate
air filter, which is designed to remove 99.999% of particles greater than 0.3 m in
size. It must be remembered that the laminar flow of air (usually vertical) will be
disturbed by the presence of any objects within the cabinet, including shielding
272
and the arms of the operator. During use, the filtered air may escape from the
front of the cabinet, when the airflow is disturbed, so operator protection cannot
be ensured.
9.7.3. Isolator cabinets
Isolator cabinets provide both operator and product protection. Figure 9.8
shows an example of a blood cell labelling isolator. The product is manipulated
through glove ports so that the interior of the cabinet is maintained totally sterile
and full operator protection is provided. Airflow within the isolator is deliberately
designed to be turbulent so that there are no dead spaces within the cabinet. The
unit illustrated incorporates a centrifuge which can be controlled externally. A
dose calibrator can be included within the isolator, so that the cell suspension
does not need to be removed from the isolator for the activity to be measured.
The isolator incorporates timed interlocks on the vacuum door seals to ensure
that the product remains sterile.
FIG.9.8. Blood cell labelling isolator (courtesy of Amercare Ltd).
273
CHAPTER 9
9.8. SHIELDING FOR RADIONUCLIDES

9.8.1. Shielding for , and positron emitters
Shielding will be required in the walls of the radiopharmacy, in any
containment enclosures, in a body shield to protect the operator at the dispensing
station, and around individual vials and syringes containing radionuclides.
Shielding of the walls of the radiopharmacy can be minimized by appropriate
local shielding around the sources being handled. Shielding may be constructed
from a variety of materials, including lead and concrete in walls, lead or tungsten
in local shielding for emitting radionuclides, and aluminium or Perspex for
pure emitters. For positron emitters, the shielding will be primarily determined
by the 511 keV annihilation photons rather than by the positrons themselves.
A low atomic number material, such as aluminium or Perspex, is used for pure
emitters since this minimizes the production of bremsstrahlung radiation. As
radiation has a finite range in materials, determined by the maximum energy,
the thickness of the shielding needs to be greater than this range to ensure that all
of the particles are absorbed. Polymethyl methacrylate (Perspex or lucite) has
a density of 1.19 g/cm3, similar to the density of tissue and water, and is highly
suitable for absorbing particles. Table 9.6 gives the maximum energy and the
range in water for four pure emitters used in nuclear medicine.
TABLE9.6. MAXIMUM ENERGY AND RANGE IN WATER FOR
FOUR EMITTERS USED CLINICALLY IN NUCLEAR MEDICINE.
THE ELECTRON RANGE HAS BEEN DETERMINED USING the
CONTINUOUS SLOWING DOWN APPROXIMATION
Emax (MeV)
Range in water (mm)
14
0.156
0.30
32
1.709
8.2
89
1.463
6.8
90
2.274
11
Radionuclide
C
P
Sr
Y
The highest surface dose rates encountered in the radiopharmacy are likely
to be from 99Mo/99mTc generators which may contain >100 GBq of 99Mo. The
primary emission from 99Mo has an energy of 740 keV, so requires several
centimetres of lead shielding to reduce the dose rates to an acceptable level. The
generator, as supplied, will already contain substantial shielding but additional
shielding will usually be required. This may be available from the generator
274
supplier specifically designed for their generator or it may be necessary to

construct or purchase an additional shield. Figure 9.9 shows a generator supplied
by ANSTO in Australia inside a dedicated lead garage. The body of the
radiochemist is shielded by the garage doors while she is attaching the shielded
elution vial prior to an elution of the generator. These shields are heavy (>20 kg),
so it is important that the bench surfaces are strong enough to support the weight.
The generator is itself quite heavy, so mechanical lifting devices may be necessary
to prevent back injuries to staff when lifting the generator into position.
FIG.9.9. Lead garage surrounding a 99Mo/99mTc generator.
Vials of radiopharmaceuticals must be kept shielded. The shields are usually

constructed so that only the rubber septum of the vial is accessible, thereby
protecting the hands of the operator during dispensing (see Fig. 9.10). The vials
themselves should never be held by the fingers once they contain radioactivity,
instead long forceps should always be used (see Fig. 9.11).
275
CHAPTER 9
FIG.9.10. Shielded vial used to hold reconstituted radiopharmaceuticals.
FIG.9.11. Using long forceps to handle a vial containing radioactivity.
276
During radiopharmaceutical preparation, dispensing and administration

to the patient, the activity is usually manipulated in syringes. The dose rates at
the surface of the syringes may exceed 1 Sv s1 MBq1 depending on the
volume of liquid and the size of the syringe. The plastic of the syringe provides
little absorption of any high energy particles, and for radionuclides used for
therapy, the surface dose rates will be in excess of 10 mSv/s, so that the annual
dose limit of 500 mSv to the extremities can easily be exceeded. Syringes should
never be more than half filled, so that the syringe can be picked up near the
plunger where the fingers are not over the activity. Syringe shields must be used
whenever possible. These must be made of Perspex for the pure emitters and
of lead or tungsten for the emitters (see Fig. 9.12). A lead glass window is
necessary to permit observation of the contents of the syringe. Syringe shields
with a spring-loaded catch to hold the syringe in place are preferable to those
using a screw, as the screw-thread wears quickly with use.
FIG.9.12. A tungsten syringe shield for emitting radionuclides and a Perspex syringe shield
for pure emitters.
277
CHAPTER 9
TABLE9.7. MEASURED TRANSMISION FACTORS FOR LEAD

Thickness
of lead (mm)
99
1.00
1.00
1.00
1.00
1.00
1.00
0.876
0.105
0.455
0.769
0.136
0.891
0.776
0.00835
0.280
0.601
0.0709
0.787
0.694
6.52 104
0.191
0.475
0.0557
0.690
0.623
5.09 10
0.135
0.379
0.0485
0.602
0.561
3.97 106
0.0983
0.306
0.0438
0.523
0.507
0.0730
0.248
0.0430
0.452
0.458
0.0551
0.203
0.0422
0.390
0.414
0.0420
0.168
0.0415
0.336
0.375
0.0324
0.139
0.0408
0.289
10
0.340
0.0253
0.116
0.0291
0.249
12
0.279
0.0157
0.0829
0.0282
0.183
14
0.229
0.0102
0.0605
0.0273
0.135
16
0.188
0.00682
0.0451
0.0193
0.0990
18
0.154
0.00476
0.0342
0.0187
0.0728
20
0.127
0.00345
0.0263
0.0132
0.0535
25
0.0774
0.00177
0.0143
0.00893
0.0247
30
0.0473
0.00104
0.00805
0.00602
0.0114
40
0.0176
4.11 104
0.00267
0.00274
0.00240
0.00124
5.00 104
50
a
Mo
0.00659
99m
67
Tc
3.10 10
131
Ga
1.71 10
201
Tla
9.04 10
511 keV
The transmission data for 201Tl includes a contribution of 1.5% of the contaminant 200Tl, the
maximum level likely to be encountered in clinical practice.
9.8.2. Transmission factors for lead and concrete

Section 1.6 indicates that the attenuation of monoenergetic photons through
materials such as lead or concrete will be exponential, characterized by the linear
attenuation coefficient or the half-value layer (HVL). However, this is only
correct for narrow beam geometries, using collimated beams of radiation, which
278
are rarely encountered in practice. Furthermore, the attenuation of the radiation

from radionuclides which emit more than one photon, such as 67Ga and 131I,
cannot be expressed as a simple HVL.
Tables 9.7 and 9.8 give the measured broad beam transmission factors for
lead and concrete for five radionuclides used in nuclear medicine and for 511 keV
photons from positron emitters. This information can be used to calculate the
required thickness of shielding around vials, for the body protection of the
operator and for the walls of the radiopharmacy. The values for 201Tl include a
contribution from 200Tl, a common contaminant, which has prominent energies at
368 keV and 1.2 MeV. A contribution of 1.5% of 200Tl has been included which is
the maximum likely value at the time of calibration.
TABLE9.8. MEASURED TRANSMISSION FACTORS FOR CONCRETE
(DENSITY: 2.35 g/cm3) (cont.)
Thickness of
concrete (mm)
99
1.00
1.00
1.00
1.00
1.00
1.00
10
0.845
0.779
0.884
0.916
0.759
0.958
20
0.718
0.607
0.769
0.825
0.581
0.909
30
0.614
0.473
0.661
0.735
0.449
0.852
40
0.527
0.368
0.564
0.649
0.349
0.789
50
0.454
0.287
0.477
0.570
0.274
0.722
60
0.393
0.224
0.402
0.498
0.217
0.653
70
0.341
0.174
0.338
0.434
0.173
0.584
80
0.296
0.136
0.282
0.377
0.139
0.518
90
0.258
0.106
0.236
0.327
0.112
0.456
100
0.225
0.0824
0.196
0.284
0.0912
0.399
120
0.172
0.0500
0.135
0.212
0.0612
0.301
140
0.132
0.0304
0.0928
0.158
0.0418
0.224
160
0.101
0.0184
0.0635
0.118
0.0290
0.166
180
0.0777
0.0112
0.0434
0.0879
0.0203
0.123
200
0.0598
0.00679
0.0296
0.0654
0.0143
0.0904
250
0.0312
0.00195
0.0113
0.0312
0.00607
0.0419
Mo
99m
Tc
67
Ga
131
201
Tla
511 keV
279
CHAPTER 9
TABLE9.8. MEASURED TRANSMISSION FACTORS FOR CONCRETE

(DENSITY: 2.35 g/cm3) (cont.)
Thickness of
concrete (mm)
300
99
Mo
0.0163
99m
67
Tc
131
Ga
5.60 104
5
0.00433
4
0.0149
201
Tla
511 keV
0.00262
4
0.0194
400
0.00443
4.61 10
500
0.00121
3.80 106 9.16 105 7.70 104 9.39 105 8.95 104
6.30 10
0.00339
4.95 10
0.00417
The transmission data for 201Tl includes a contribution of 1.5% of the contaminant 200Tl, the
maximum level likely to be encountered in clinical practice.
9.9. DESIGNING A RADIOPHARMACY

Every radiopharmacy is unique and there is no one design that can be used
in all situations. The requirements of a single camera practice using only 99mTc
radiopharmaceuticals will be very different from a large teaching hospital with
PET facilities and in-patient radionuclide therapy rooms. However, in addition
to the general building requirements given in section 3.1.3 of Ref. [9.10], there
are some general rules specific to a radiopharmacy that can be applied in most
situations:
The radiopharmacy should be located in an area that is not accessible to
members of the public.
There should be easy access from the radiopharmacy to the injection rooms
and imaging rooms to minimize the distance that radioactive materials need
to be transported.
The radiopharmacy should not be adjacent to areas that require a low and
constant radiation background such as a counting room.
There should be an area within the radiopharmacy designated as a
non-active area that is used for record keeping and/or computer entry.
A refrigerator will be required for the storage of lyophilized
radiopharmaceutical kits. A laboratory-grade unit is preferred to ensure that
the temperature remains constant.
A dedicated dispensing area with a body shield and lead glass viewing
window will be required. This will normally be adjacent to the dose
calibrator, so that the dispensed activity can be measured while the operator
is still protected by the body shield. The thickness of the shield and window
will depend on the radionuclide or radionuclides in use. PET radionuclides
280
will require substantial thickness and lead glass should be supplied as a

single block rather than as a stack of thinner sheets.
A storage area will be required for reconstituted radiopharmaceuticals, in
shielded containers, together with radiopharmaceuticals purchased ready
for dispensing such as 67Ga-citrate and 201Tl-chloride.
The radiopharmacy must contain facilities for radioactive waste disposal.
This will normally include separate shielded storage bins for short lived
radionuclides such as 99mTc and for radionuclides with longer half-lives
such as 131I. In addition, there must be shielded containers for sharps, such
as syringes with needles. A separate shielded storage bin may be required
if a large number of bulky items, such as aerosol or Technegas kits, need to
be stored.
If a Mo/Tc generator is used, this should be positioned away from the
dispensing area to minimize the dose received by the person dispensing the
radiopharmaceuticals. Some countries require the generator to be housed
inside a laminar flow cabinet. All local regulatory requirements must be
taken into account when designing the radiopharmacy.
If cell labelling procedures are to be performed, a dedicated area with a
laminar flow cabinet or isolator will be required to ensure that the product
remains sterile during the labelling procedure.
A fume cupboard, together with an activated charcoal filter on the exhaust,
will be required if radioiodination procedures are to be performed.
Some radiopharmaceuticals require a heating step in their preparation.
This is often performed using a temperature controlled heating block. This
must be in a dedicated separately shielded area, particularly as several
gigabecquerels of 99mTc are often involved. Similarly, the radiolabelling of
blood samples may require local shielding of mixers and centrifuges.
Wall, floor and ceiling surfaces should be smooth, impervious and durable,
and free of externally mounted features such as pipes or ducts to facilitate
any radioactive decontamination.
Bench surfaces should be constructed of plastic laminate or resin composites
or stainless steel, and benches must be able to safely withstand the weight
of any required lead shielding.
Hand washing facilities must be available which can be operated without
the use of the operators hands to prevent the spread of any contamination.
An eye-wash should also be available.
A contamination monitor must be available in a readily accessible location.
A wall-mounted monitor to check for any hand contamination should be
mounted near the exit from the radiopharmacy. A model which can be
removed and used as a general contamination monitor is useful.
281
CHAPTER 9
9.10. SECURITY OF THE RADIOPHARMACY

Until relatively recently, the safety of the staff when handling and storing
radioactive materials was the sole concern when designing a radiopharmacy. The
security of the radioactivity was often not specifically addressed. Unfortunately,
it is now apparent that radioactive materials can be used for malicious purposes
and the security of the radiopharmacy must now be considered.
The IAEA has categorized radioactive sources on a scale of 1 to 5, based
on activity and nuclide, where category 1 is potentially the most hazardous.
Sources categorized as 1, 2 or 3 are known as security enhanced sources. The
security measures in place for safety purposes are considered adequate to ensure
the physical security of category 4 and 5 sources. Legislation is now, or will be,
in place in each jurisdiction to address the security of security enhanced sources.
This currently only applies to sealed sources, and no sealed sources used in
nuclear medicine are categorized as either 1, 2 or 3. However, the principles can
be applied to unsealed sources. A Mo/Tc generator with an activity of greater
than 300 GBq is a category 3 source.
Radioactive materials are at most risk of being stolen or lost when they
are being transported to and from the facility. They will be in the appropriate
transport container and, therefore, can be easily handled by someone with
malicious intent. It is essential that all consignments of radioactive materials to
the nuclear medicine facility are left in a secure area and not left, for example,
on a loading dock. During working hours, all deliveries must be signed for by a
designated staff member and the material safely unpacked and stored within the
department. Some deliveries may occur outside of working hours. In this case,
a dedicated secure area must be provided where the radioactive materials can
be left. A key could be provided to the supplier for this area only, so that the
radioactivity can be safely and securely delivered, but access to other parts of the
facility is prevented. The supplier may need to be accompanied by the facilitys
security staff when delivering the shipment.
Whether secure access (such as electronic card access) to the radiopharmacy
during working hours is required will depend on local requirements and the
layout of the nuclear medicine department. It is essential that only trained nuclear
medicine staff have access to the radiopharmacy. The need for controlled access
needs to be balanced against the possibility of inadvertent contamination of the
door or access mechanism by staff returning to the radiopharmacy.
282
9.11. RECORD KEEPING

The local regulations may specify the minimum records that must be kept
at the facility, the form in which these must be kept (paper and/or electronic)
and the time for which the records must be kept. Records can be generated as
part of the quality assurance (QA) programme, for the receipt and subsequent
administration of a radiopharmaceutical to a patient, and for waste disposal.
9.11.1. Quality control records
A key element of any QA programme is proper record keeping, so that
any long term trends associated with a particular item of equipment or batch of
radiopharmaceuticals can be identified and acted on before image quality and/or
patient dose are compromised. Records should, at the very least, include details
of:
Acceptance testing of the dose calibrator;
All constancy tests;
Radiopharmaceutical testing.
Failures identified at acceptance or constancy testing and
radiopharmaceutical testing, and the actions taken to remedy those failures,
should be documented and these records kept for the lifetime of the equipment.
The following records should be kept for all generator elutions:
Time of elution;
Volume of eluate;
Technetium-99m activity;
Molybdenum-99 activity;
Radionuclidic purity.
9.11.2. Records of receipt of radioactive materials
Complete records of the radionuclide, activity, chemical form, supplier,
suppliers batch number and purchase date should be kept. On arrival, if a
package containing radioactive material is suspected of being damaged, the
package should be:
Monitored for leakage with a wipe test;
Checked with a survey meter for unexpectedly high external radiation
levels.
283
CHAPTER 9
If a package is damaged or suspected of being damaged, the supplier should

be contacted immediately, and the details recorded.
9.11.3. Records of radiopharmaceutical preparation and dispensing
The preparation of radiopharmaceuticals needs to be performed in
accordance with the manufacturers requirements as specified in the product
documentation, including any quality control such as thin-layer chromatography.
Records of each preparation should include the:
Name of the radiopharmaceutical;
Cold kit batch number;
Date of manufacture;
Batch number of final product;
Radiochemical purity results;
Expiry date.
A record for each patient dose dispensed must be kept with the:
Name of the patient;
Name of the radiopharmaceutical;
Measured radioactivity;
Time and date of measurement.
All unit patient doses (syringes, capsules or vials) supplied by a central
radiopharmacy should identify the patients name and the radionuclide and
radiopharmaceutical form. These should be verified on arrival and the activity
should be confirmed in a dose calibrator prior to administration to the patient,
and recorded as above.
9.11.4. Radioactive waste records
Radioactive waste generated within a nuclear medicine facility usually
consists of radionuclides with half-lives of less than one month. This waste will
normally be stored on-site, be allowed to decay to background radiation levels
and then be disposed of as normal waste or biologically contaminated waste
(see Section 3.4.7). It is, therefore, not normally necessary to keep records of
radioactive waste disposal from the facility, but it will be necessary to keep
records of the waste in storage while it decays. In some circumstances, the waste
will contain a single known radionuclide, such as 131I from patients receiving
radioiodine ablation therapy. In many cases, the waste will contain a mixture of
284
short lived radionuclides. Each package of waste (bag, sharps container, wheeled
bin) must be marked with the:
Radionuclide, if known;
Maximum dose rate at the surface of the container or at a fixed distance
(e.g. 1 m);
Date of storage.
The above information should be recorded, together with information
identifying the location of the container within the store, and the likely release
date (e.g. ten half-lives of the longest lived radionuclide in the container).
When the package is finally released for disposal, the record should be
updated to record the dose rate at that time, which should be at background levels,
the date of disposal, and the identification of the person authorizing its disposal.
Old sealed sources previously used for quality control or transmission
scans, such as 137Cs, 57Co, 153Gd and 68Ge, should be kept in a secure store until
the activity has decayed to a level permitted for disposal, or the source can be
disposed of by a method approved by the regulatory authority.
REFERENCES
[9.1] TYLER, D.K., WOODS, M.J., Syringe calibration factors for the NPL Secondary
Standard Radionuclide calibrator for selected medical radionuclides, Appl. Radiat.
Isot. 59(2003) 367372.
[9.2] INTERNATIONAL ELECTROTECHNICAL COMMISSION, Calibration and Usage
of Ionization Chamber Systems for Assay of Radionuclides, IEC 61145:1992, IEC
(1992).
[9.3] INTERNATIONAL ELECTROTECHNICAL COMMISSION, Medical Electrical
Equipment Radionuclide Calibrators Particular Methods for Describing
Performance, IEC 61303:1994, IEC (1994).
[9.4] INTERNATIONAL ELECTROTECHNICAL COMMISSION, Nuclear Medicine
Instrumentation Routine Tests Part 4: Radionuclide Calibrators, IEC/TR
61948-4:2006, IEC (2006).
[9.5] American National Standards Institute, Calibration and Usage of Dose Calibrator
Ionization Chambers for the Assay of Radionuclides, ANSI N42.13-2004, ANSI
(2004).
[9.6] INTERNATIONAL COMMISSION ON RADIOLOGICAL PROTECTION,
Radiation Dose to Patients from Radiopharmaceuticals, Publication 53, Pergamon
Press, Oxford and New York (1988).
285
CHAPTER 9

Radiation Dose to Patients from Radiopharmaceuticals (Addendum to ICRP
Publication 53), Publication 80, Pergamon Press, Oxford and New York (1998).
Radiation Dose to Patients from Radiopharmaceuticals Addendum 3 to ICRP
Publication 53, Publication 106, Elsevier (2008).
Radiological Protection and Safety in Medicine, Publication 73, Pergamon Press,
Oxford and New York (1996).
[9.10] INTERNATIONAL ATOMIC ENERGY, Applying Radiation Safety Standards in
Nuclear Medicine, Safety Reports Series No. 40, IAEA, Vienna (2005).
[9.11] EUROPEAN ALARA NETWORK (2007),
http://www.eu-alara.net/index.php?option=com_content&task=view&id=156&
Itemid=53
[9.12] INTERNATIONAL COMMISSION ON RADIOLOGICAL PROTECTION, Limits
for Intakes of Radionuclides by Workers, Publication 30, Pergamon Press, Oxford and
New York (1979).
[9.13] INTERNATIONAL COMMISSION ON RADIOLOGICAL PROTECTION, 1990
Recommendations of the International Commission on Radiological Protection,
Publication 60, Pergamon Press, Oxford and New York (1991).
[9.14] INTERNATIONAL COMMISSION ON RADIOLOGICAL PROTECTION, Annual
Limits on Intake of Radionuclides by Workers Based on the 1990 Recommendations,
Publication 61, Pergamon Press, Oxford and New York (1991).
BIBLIOGRAPHY
GADD, R., et al., Protocol for Establishing and Maintaining the Calibration of Medical
Radionuclide Calibrators and their Quality Control, Measurement Good Practice Guide No. 93,
National Physical Laboratory, UK (2006).
GROTH, M.J., Empirical dose rate and attenuation data for radionuclides in nuclear medicine,
Australas. Phys. Eng. Sci. Med. 19 (1996) 160167.
NATIONAL HEALTH AND MEDICAL RESEARCH COUNCIL (Australia), Recommended
Limits on Radioactive Contamination on Surfaces in Laboratories, Radiation Health Series
No. 38, NHMRC (1995).
SCHRADER, H., Activity Measurements with Ionization Chambers, Monographie Bureau
International des Poids et Mesures No. 4 (1997).
286
CHAPTER 10
NON-IMAGING DETECTORS AND COUNTERS
P.B. ZANZONICO
Memorial Sloan Kettering Cancer Center,
New York, United States of America
10.1. INTRODUCTION
Historically, nuclear medicine has been largely an imaging based
specialty, employing such diverse and increasingly sophisticated modalities as
rectilinear scanning, (planar) gamma camera imaging, single photon emission
computed tomography (SPECT) and positron emission tomography (PET).
Non-imaging radiation detection, however, remains an essential component of
nuclear medicine. This chapter reviews the operating principles, performance,
applications and quality control (QC) of the various non-imaging radiation
detection and measurement devices used in nuclear medicine, including survey
meters, dose calibrators, well counters, intra-operative probes and organ uptake
probes. Related topics, including the basics of radiation detection, statistics of
nuclear counting, electronics, generic instrumentation performance parameters
and nuclear medicine imaging devices, are reviewed in depth in other chapters
of this book.
10.2. OPERATING PRINCIPLES OF RADIATION DETECTORS
Radiation detectors encountered in nuclear medicine may generally
be characterized as either scintillation or ionization detectors (Fig. 10.1). In
scintillation detectors, visible light is produced as radiation excites atoms of a
crystal or other scintillator and is converted to an electronic signal, or pulse, and
amplified by a photomultiplier tube (PMT) and its high voltage (5001500 V).
In ionization detectors, free electrons produced as radiation ionizes a stopping
material within a sensitive volume are electrostatically collected by a bias voltage
(10500 V) to produce an electron signal. In both scintillation and ionization
detectors, the unprocessed signal is then shaped and amplified. For some types
of detector, the resulting pulses are sorted by their amplitude (or pulse height),
which is related to the X ray or ray energy absorbed in the detector.
287
CHAPTER 10
10.2.1. Ionization detectors

Detector materials in the most common ionization detectors are gaseous
and such detectors are, therefore, often known as gas filled detectors; however,
as discussed in the following, solid state ionization detectors also exist. The
two most widely encountered gas ionization detectors in nuclear medicine are
dose calibrators and Geiger counters. The principal difference between these
detectors is the magnitude of the bias voltage between the anode and cathode, as
indicated graphically in Fig. 10.2. When the bias voltage is less than 300 V, ion
pairs (i.e. free electrons and positive ions) produced as radiation passes through
the sensitive volume may recombine, thereby preventing at least some electrons
from reaching the anode and yielding an artefactually low signal. The 0300 V
range is, therefore, called the recombination region.
Pulses in a preset pulse height range are counted
Pulse sorted by amplitude (height)
Pulse shaped and amplified
High voltage
5001500 V
Unprocessed electron signal

(pulse)
Photomultiplier
tube
Photocathode
Detector
material
Anode
Bias voltage
10500 V
Cathode
X or ray
Crystal
X or ray
(a) Scintillation detector
(b) Ionization detector
FIG.10.1. Basic design and operating principles of (a) scintillation and (b) ionization
detectors.
At a bias voltage of 300 V, all of the primary electrons (i.e. the electrons
produced directly by ionization of the detector material by the incident radiation)
are collected at the anode and the detector signal is, thereby, maximized. Since
288
300
Spontaneous discharge region
Geiger counter region
103
Proportional counter region
106
Ionization chamber region
109
Recombination region
Signal (total number of electrons per primary electron)
600
900
Bias voltage (V)
1200
FIG.10.2. The signal (expressed as the amplification factor, that is, the total number of
electrons per primary electron produced in the detector material) as a function of the bias
voltage for gas filled ionization detectors. The principal difference among such detectors is the
magnitude of the bias voltage between the anode and cathode. The amplification factors and
the voltages shown are approximate.
there are no additional primary electrons to collect, increasing the bias voltage
further (up to 600 V) does not increase the signal. The 300600 V range,
where the overall signal is equivalent to the number of primary electrons and,
therefore, proportional to the energy of the incident radiation, is called the
ionization chamber region. At a bias voltage of 600900 V, however, the large
electrostatic force of attraction of the anode accelerates free electrons, as they
travel towards the anode, to sufficiently high speeds to eject additional orbital
electrons (i.e. secondary electrons) within the sensitive volume, contributing to
an increasing overall signal the higher the voltage, the more energetic the
electrons and the more secondary electrons are added to the overall signal. The
number of electrons comprising the overall signal is, thus, proportional to the
289
CHAPTER 10
primary number of electrons and the energy of the incident radiation, and the
600900 V range is, therefore, called the proportional counter region. As the bias
voltage is increased further, beyond 900 V (up to 1200 V), free electrons (primary
and secondary) are accelerated to very high speeds and strike the anode with
sufficient energy to eject additional electrons from the anode surface itself. These
tertiary electrons are, in turn, accelerated back to the anode surface and eject even
more electrons, effectively forming an electron cloud over the anode surface and
yielding a constant overall signal even with further increase in the bias voltage.
The 9001200 V range is called the Geiger counter (or GeigerMller) region.
Importantly, the magnitude of the charge represented by this electron cloud is
independent of the number of electrons initiating its formation. Therefore,
in contrast to ionization chamber and proportional counter signals, the Geiger
counter signal is independent of the energy of the incident radiation. Finally,
beyond a bias voltage of 1200 V, atoms within the detector material are ionized
even in the absence of ionizing radiation (i.e. undergo spontaneous ionization),
producing an artefactual signal; the voltage range beyond 1200 V is known as the
spontaneous discharge region.
Although the bias voltage is the principal difference among different types
of gas filled ionization detectors, there may be other differences. The sensitive
volume, for example, may or may not be sealed. Unsealed sensitive volumes
contain only air at atmospheric (ambient) pressure. For detectors with unsealed
volumes, the signal must be corrected by calculation for the difference between
the temperature and pressure at which the detector was calibrated (usually
standard temperature and pressure: 27C and 760 mm Hg, respectively) and
the ambient conditions at the time of an actual measurement. For detectors with
sealed volumes, gases other than air (e.g. argon) may be used and the gas may be
pressurized, providing higher stopping power, and, therefore, higher sensitivity,
than detectors having a non-pressurized gas in the sensitive volume. In addition,
different geometric arrangements of the anode and cathode, such as parallel
plates (used in some ionization chambers), a wire along the axis of a cylinder
(used in Geiger counters), etc., may be used.
The functional properties and, therefore, the applications of the various
types of ionization detector ionization chambers, proportional counters and
Geiger counters are largely dictated by their respective bias voltage dependent
signal (Table 10.1). Ionization chambers are widely used in radiation therapy to
calibrate the output of therapy units and in nuclear medicine as dose calibrators
(i.e. devices used to assay radiopharmaceutical activities). The relatively
low sensitivity of ionization chambers is not a major disadvantage for such
applications, as the radiation intensities encountered are typically rather large.
The stability of the response is an important advantage, however, as it allows
the use of unconditioned AC electrical power (i.e. as provided by ordinary wall
290
outlets). Proportional counters, because of their need for a stable bias voltage
and, therefore, specialized power supplies, are restricted to research applications
(e.g. in radiobiology) where both higher sensitivity and the capability of energy
discrimination may be advantageous. Proportional counters often employ an
unsealed, gas flow-through sensitive volume. Geiger counters, because of their
high sensitivity and stability with respect to voltage (allowing the use of a portable
power supply such as an ordinary battery), are widely used as survey meters to
measure ambient radiation levels and to detect radioactive contamination. For
such applications, sensitivity, and not energy discrimination, is critical. As with
dose calibrators, Geiger counters have sealed sensitive volumes, avoiding the
need for temperaturepressure corrections.
In addition to the more familiar gas filled ionization detectors, solid state
ionization detectors are now available. Such detectors are based on a family of
materials known as semiconductors. The pertinent difference among (crystalline)
solids conductors, insulators and semiconductors is related to the widths of
their respective electron forbidden energy gaps. In a semiconductor, the highest
energy levels occupied by electrons are completely filled but the forbidden
gap is narrow enough (<2 eV) to allow radiative or even thermal excitation at
room temperature, thereby allowing a small number of electrons to cross the
gap and occupy energy levels among the otherwise empty upper energy levels.
Such electrons are mobile and, thus, can be collected by a bias voltage, with the
amplitude of the resulting signal being equivalent to the number of electrons
produced by the radiation and, therefore, proportional to the radiation energy.
Although many semiconductor materials have suitably large energy gaps (~2 eV),
techniques must be available to produce crystals relatively free of structural
defects. Defects (i.e. irregularities in the crystal lattice) can trap electrons
produced by radiation and, thus, reduce the total charge collected, degrading
the sensitivity and overall detector performance of semiconductors. Practical,
reasonably economical crystal growing techniques have been developed
for cadmium telluride (CdTe), cadmium zinc telluride (CZT) and mercuric
iodide (HgI2), and these detectors have been incorporated into commercial
intra-operative gamma probes and, on a limited basis, small field of view gamma
cameras.
291
CHAPTER 10
TABLE10.1. PROPERTIES OF GAS FILLED IONIZATION DETECTORS

Ionization
detector
Proportional
counter
Geiger
counter
300600 V
600900 V
9001200 V
Response stable with respect to

voltage?a
Yes
No
Yes
Sensitivityb
Low
Intermediate
High
Capable of energy
discrimination?c
Yes
Yes
No
Dose calibrator
Research
Survey meter
Bias voltage operating range
Applications
a
b
c
The stability with respect to the bias voltage corresponds to a constant signal over the
respective detectors operating voltage range. In contrast to ionization detectors and Geiger
counters, proportional counters are unstable with respect to the bias voltage and, thus,
require specialized, highly stable voltage sources for constancy of response.
The sensitivity of a detector is related to its amplification factor (see Fig. 10.2).
If the total number of electrons comprising the signal is proportional to the number of
electrons directly produced by the incident radiation and, therefore, proportional to its
energy, as in ionization detectors and proportional counters, radiations of different energies
can be discriminated (i.e. separated) on the basis of the signal amplitude.
10.2.2. Scintillation detectors

In scintillation detectors, radiation interacts with and deposits energy in a
scintillator, most commonly, a crystalline solid such as thallium-doped sodium
iodide (NaI(Tl)). The radiation energy thus deposited is converted to visible light.
As the light is emitted isotropically (i.e. in all directions), the inner surface of
the light-tight crystal housing is coated with a reflective material so that light
emitted towards the sides and front of the crystal are reflected back towards a
PMT (Fig. 10.3); this maximizes the amount of light collected and, therefore,
the overall sensitivity of the detector. Interposed between the back of the crystal
and the entrance window of the PMT is the light pipe, nowadays simply a thin
layer of transparent optical gel. The light pipe optically couples the crystal to the
PMT and, thus, maximizes the transmission (>90%) of the light signal from the
crystal into the PMT. When struck by light from the crystal, the photocathode
coated on the inner surface of the PMT emits electrons. Immediately beyond the
photocathode (which is at ground, that is, 0 V) is the focusing grid, maintained at
a relatively low positive voltage on the order of 10 V. As electrons pass through
the focusing grid, they are attracted by a relatively large positive voltage, ~300 V,
on the first of a series of small metallic elements called dynodes. The resulting
high speed impact of each electron results in the ejection from the dynode surface
of an average of three electrons. The electrons thus ejected are then attracted by
292
the even larger positive voltage, ~400 V, on the second dynode. The impact of
these electrons on the second dynode surface ejects an additional three electrons,
on average, for each incident electron. Typically, a PMT has 1012 such dynodes
(or stages), each ~100 V more positive than the preceding dynode, resulting in
an overall electron amplification factor of 310312 for the entire PMT. At the last
anode, an output signal is generated. The irregularly shaped PMT output signal
is then shaped by a preamplifier and further amplified into a logic pulse that
can be further processed electronically. The resulting electrical pulses, whose
amplitudes (or heights) are proportional to the number of electrons produced
at the PMT photocathode are, therefore, also proportional to the energy of the
incident radiation. These pulses can then be sorted according to their respective
heights by an energy discriminator (also known as a pulse height analyser) and
those pulses with a pulse height (i.e. energy) within the preset photopeak energy
window (as indicated by the pair of dashed horizontal lines overlying the pulses
in Fig. 10.3) are counted by a timer/scaler.
Advantageous features of scintillation detectors include:
High electron density (determined by mass density and effective atomic
number Zeff);
High light output;
For certain applications such as PET, speed of light emission.
High mass density and effective atomic number maximize the crystal
stopping power (i.e. linear attenuation coefficient ) and, therefore, sensitivity.
In addition, a higher atomic number crystal will have a higher proportion of
photoelectric than Compton interactions, thus facilitating energy discrimination
of photons which underwent scatter before entering the crystal. High light output
reduces statistical uncertainty (noise) in the scintillation and associated electronic
signal and, thus, improves energy resolution and scatter rejection. Other detector
considerations include:
Transparency of the crystal to its own scintillations (i.e. minimal
self-absorption);
Matching of the index of refraction of the crystal to that of the
photodetector (specifically, the entrance window ( 1.5) of the PMT);
Matching of the scintillation wavelength to the light response of the
photodetector (the PMT photocathode, with maximum sensitivity in the
390410 nm, or blue, wavelength range);
Minimal hygroscopic behaviour.
293
CHAPTER 10
To date, the most widely used scintillators in nuclear medicine include:

NaI(Tl), bismuth germanate (BGO), cerium-doped lutetium oxyorthosilicate
(LSO(Ce) or LSO) and cerium-doped gadolinium oxyorthosilicate (GSO(Ce) or
GSO). NaI(Tl) is used in cameras/SPECT systems, well counters and organ
uptake probes, and remains the most widely used scintillator in clinical practice;
BGO, LSO and GSO are the scintillators of choice in PET scanners because
of their higher stopping power for the 511 keV positronnegatron annihilation
photons. Thallium- and sodium-doped caesium iodide (CsI(Tl) and CsI(Na),
respectively) and cadmium tungstate as well as NaI(Tl), BGO and LSO have also
been used in intra-operative probes.
10.3. RADIATION DETECTOR PERFORMANCE
Radiation detectors may be quantitatively characterized by many different
performance parameters, particularly for those detectors such as cameras which
localize (image) as well as count radiation. For non-imaging radiation detectors
and counters, however, the most important performance parameters are sensitivity
(or efficiency), energy resolution and count rate performance (or speed).
10.3.1. Sensitivity
Sensitivity (or efficiency) is the detected count rate per unit activity (e.g. in
counts per minute per megabecquerel). As the count rate detected from a given
activity is highly dependent on the sourcedetector geometry and intervening
media, characterization of sensitivity can be ambiguous. There are two distinct
components of overall sensitivity, geometric sensitivity and intrinsic sensitivity.
Geometric sensitivity is the fraction of emitted radiations which intersect, or strike,
the detector, that is, the fraction of the total solid angle subtended at the detector
by the source. It is, therefore, directly proportional to the radiation-sensitive
detector area and, for a point source, inversely proportional to the square of the
sourcedetector distance. Intrinsic sensitivity is the fraction of radiation striking
the detector which is stopped within the detector. Intrinsic sensitivity is directly
related to the detector thickness, effective atomic number and mass density, and
decreases with increasing photon energy, since higher energy photons are more
penetrating and are more likely to pass through a detector without interacting.
Characteristic X rays and rays are emitted from radioactively decaying
atoms with well defined discrete energies. Even in the absence of scatter, however,
output pulses from absorption of these radiations will appear to originate over a
range of energies, reflecting the relatively coarse energy resolution of the detector.
For this reason, many radiation detectors employ some sort of energy-selective
294
counting using an energy range, or window, such that radiations are only counted
if their detected energies lie within that range (Figs. 10.3 and 10.4(a)). At least
for scintillation detectors, a so-called 20% photopeak energy window, E 10%
of E, (e.g. 126154 keV for the 140 keV ray of 99mTc) is employed, where E
is the photopeak energy of the X ray or ray being counted. For such energyselective counting, overall sensitivity appears to increase as the photopeak energy
window is widened. However, this results in acceptance of more scattered as well
as primary (i.e. unscattered) radiations.
For each radionuclide and energy window (if applicable) for which a
particular detector is used, the detector should be calibrated, that is, its sensitivity
(e.g. in cpm/MBq) S determined, at installation and periodically thereafter:
S=
Rg R b
A 0e t
(10.1)
where
Rg is the gross (i.e. total) count rate (cpm) of the radionuclide source (RS);
Rb is the background (BG), or blank, count rate (cpm);
A0 is the activity (MBq) of the radionuclide source at calibration;
is the physical decay constant (in month1 or a1, depending on the half-life)
of the calibration radionuclide;
and t is the time interval (in months or years, respectively, again depending
on the half-life) between the calibration of the radionuclide and the current
measurement.
As noted, sensitivity is highly dependent on the sourcedetector counting
geometry (including the size and shape of the source and the sourcedetector
distance), and the measured value, thus, applies exactly only for the geometry
used for the measurement.
10.3.2. Energy resolution
Energy resolution quantifies the ability of a detector to separate, or
discriminate, radiations of different energies. As illustrated in Fig. 10.4(b),
energy resolution is generally given by the width of the bell shaped photopeak,
specified as the full width at half maximum (FWHM = E) height expressed as
a percentage of the photopeak energy E, FWHM (%) =
E
100%. It is related
E
295
CHAPTER 10
Output
signal
Photomultiplier tube
Anode
+ 1200 V
106:1e
+ 1000 V
+ 600 V
Dynodes
Photocathode
Entrance
window
Reflective
inner surface
of crystal housing
Amplifier
+ 1100 V
Magnetic
shielding
Focusing
grid
Preamplifier
+ 500 V
High
voltage
supply
106
+ 400 V
+ 300 V
Energy
discriminator
Energy E0
Timer/scaler
Light
photon
Light pipe
Scintillator
crystal
Display
X or ray
FIG.10.3. The basic design and operating principle of photomultiplier tubes and scintillation
detectors.
to the Poisson noise, or statistical uncertainty, inherent in the detection process.

The importance of energy resolution lies in scatter rejection, particularly for
imaging detectors. Radiation loses energy when undergoing Compton scatter
within the source and the lower energy scattered radiations may, therefore, be
discriminated from the primary radiations. However, the finite energy resolution
of radiation detectors (i.e. the width of the photopeak in the energy spectrum)
means that there will be overlap of scattered and primary radiations, as illustrated
in Fig. 10.4(a). As energy resolution improves (i.e. the FWHM (%) decreases and
the photopeak becomes narrower), the separation of unscattered and scattered
radiations increases and more counts corresponding to scattered radiation may be
eliminated, while discarding fewer counts corresponding to unscattered radiation.
10.3.3. Count rate performance (speed)
Radiation detectors have a finite dead time or pulse resolving time
typically 510 s for modern scintillation detectors and associated count
losses. The dead time is the length of time required for a counting system to
296
20% photopeak
energy window
Relative number of counts
(a)
10
8
6
4
Primary photons
Scattered photons
Total photons
2
60
80 100 120
Photon energy E (keV)
140
137Cs
160
80
Maximum height
60
FWHM(%) =
40
DE
46
=
100 = 7%
Eg
662
20
100
200
300
400
500
600
Photon energy E (keV)
E = 46 keV
Relative number of counts
100
40
662 keV
20
(b)
140 keV
99mTc
FWHM
maximum
height
700
FIG.10.4. (a) Energy spectrum for the 662 keV rays emitted by 137Cs, illustrating the
definition of energy resolution as the percentage full width at half maximum (FWHM) of the
photopeak energy E. (b) Energy spectrum for the 140 keV rays emitted by 99mTc, illustrating
the contributions of primary (unscattered) and scattered radiation counts. In (a) and (b), the
energy spectra were obtained with a thallium-doped sodium iodide (NaI(Tl)) scintillation
detector.
record an event, during which additional events cannot be recorded. As a result,

the measured count rate is lower than the actual count rate. Radiation detectors
are characterized in terms of count rate performance as either non-paralysable
or paralysable (Fig. 10.5). In non-paralysable systems, only radiation which is
actually counted prevents the counting of subsequent radiation interacting with
the detector during the dead time of that preceding radiation. In a paralysable
297
CHAPTER 10
detector, however, even radiation which is not counted (i.e. which interacts
with the detector during the dead time of a previous event) prevents counting
of subsequent incoming radiations during the time interval corresponding to
its dead time. Geiger counters (with quenching gas) behave as non-paralysable
systems but most detectors, including scintillation detector based systems,
such as well counters, cameras and PET scanners, are paralysable. Modern
scintillation detectors generally incorporate automated algorithms to yield count
rates corrected for dead time count losses.
Observed count rate
Ideal: no dead time

count losses
Non-paralysable
Paralysable
True count rate
FIG.10.5. The observed versus true count rates for paralysable and non-paralysable radiation
detectors. For paralysable detectors, the observed count rate increases to a maximum value
with increasing true count rate (e.g. with increasing activity) and then decreases as the true
count rate is further increased. For non-paralysable detectors, the observed count rate also
increases with increasing true count rate, asymptotically approaching a maximum value as the
true count rate is further increased. In both cases, the maximum observed count rate is directly
related to the detectors dead time .
10.4. DETECTION AND COUNTING DEVICES

10.4.1. Survey meters
Survey meters, an essential component of any radiation safety programme,
are portable, battery operated, gas filled ionization detectors (or, to a much
more limited extent, solid state scintillation detectors) used to monitor ambient
radiation levels, that is, exposure rates (e.g. in coulombs per kilogram of air per
hour (C kg1 h1)) or count rates (e.g. in cpm). Among ionization detector survey
meters, so-called cutie-pies are relatively low sensitivity ionization chambers
298
(i.e. are operated at a relatively low potential difference between the anode
and cathode) and are designed for use where relatively high fluxes of X rays
and rays are encountered. The more familiar Geiger counters are operated at
a high potential difference (Fig. 10.2), providing a high electron amplification
factor and, thus, high sensitivity. Geiger counters are, therefore, well suited for
low level surveys, for example, checking for radioactive contamination. Both
cutie-pies and Geiger counters are generally calibrated in terms of exposure rate.
As an ionization chamber, the cutie-pies electron signal depends on the energy
of the detected X rays or rays and is, therefore, directly related to the exposure
for all radionuclides. For Geiger counters, on the other hand, signal pulses have
the same amplitude regardless of the energy of the incoming radiation. Thus,
Geiger counter calibration results apply only to the particular radionuclide(s)
used to calibrate the counter (see below). Solid state detectors employ a
non-air-equivalent crystal as the detection medium and, thus, cannot measure
exposure rates, only count rates.
10.4.2. Dose calibrator
The dose calibrator, used for assaying activities in radiopharmaceutical
vials and syringes and in other small sources (e.g. brachytherapy sources),
is a pressurized gas filled ionization chamber with a sealed sensitive volume
configured in a well-type geometry. While the intrinsic sensitivity of the dose
calibrator, as that of other gas filled detectors, is relatively low, the well-type
configuration of its sensitive volume provides high geometric efficiency1, making
the overall sensitivity entirely adequate for the relatively high radiopharmaceutical
activities (of the order of 10100 MBq) typically encountered in clinical nuclear
medicine. Dose calibrators are equipped with isotope specific push-buttons
and/or a potentiometer (with isotope-specific settings provided) to adjust for
differences in energy dependent response and to thereby yield accurate readouts
of activity (i.e. kBq or MBq) for any radioisotope.
10.4.3. Well counter
Well counters are used for high sensitivity counting of radioactive
specimens such as blood or urine samples or wipes from surveys of removable
contamination (i.e. wipe testing). Such counting results can be expressed in
1
The solid angle subtended at the centre of a sphere by the total surface of the sphere
is 4 steradians; a steradian is the unit of solid angle. A well-type detector configuration
approximates a point source completely surrounded by a detector, yielding a per cent geometric
efficiency of 100%, and is, therefore, referred to as a 4 counting geometry.
299
CHAPTER 10
terms of activity (e.g. MBq) using the measured isotope specific calibration
factor (cpm/MBq) (see Eq. (10.1)). Such devices are generally comprised of a
cylindrical scintillation crystal (most commonly, NaI(Tl)) with a circular bore
(well) for the sample drilled part-way into the crystal and backed by a PMT and
its associated electronics. An alternative design for well counters is the so-called
through-hole detection system in which the hole is drilled through the entire
crystal. The through-hole design facilitates sample exchange, and because
samples are centred lengthwise in the detector, yields a more constant response
for different sample volumes as well as slightly higher sensitivity than the well
counters. In both the well and through-hole designs, the crystal is surrounded by
thick lead shielding to minimize the background due to ambient radiation.
Scintillation counters are often equipped with a multichannel analyser
for energy (i.e. isotope) selective counting and an automatic sample changer
for automated counting of multiple samples. Importantly, because of their high
intrinsic and geometric efficiencies (resulting from the use of a thick crystal and
a well-type detector configuration, respectively), well counters are extremely
sensitive and, in fact, can reliably be used only for counting activities up to
~100 kBq; at higher activities, and even with dead time corrections applied,
dead time counting losses may still become prohibitive and the measured counts
inaccurate. Modern well counters often include an integrated computer which is
used to create and manage counting protocols (i.e. to specify the isotope, energy
window, counting interval, etc.), manage sample handling, and apply background,
decay, dead time and other corrections, and, thus, yield dead time-corrected net
count rate decay corrected to the start of the current counting session.
10.4.4. Intra-operative probes
Intra-operative probes (Fig. 10.6), small hand-held counting devices,
are now widely used in the management of cancer, most commonly to more
expeditiously identify and localize sentinel lymph nodes and, thereby, reduce
the need for more extensive surgery as well as to identify and localize visually
occult disease at surgery following systemic administration of a radiolabelled
antibody or other tumour-avid radiotracer. Although intra-operative probes have
been used almost exclusively for counting X rays and rays, beta (electron and
positron) probes constructed with plastic scintillators have also been developed.
In addition, small (~10 cm) field of view intra-operative cameras have recently
become available. Intra-operative probes are available with either scintillation
or semiconductor (ionization) detectors. Scintillation detector based probes have
the advantages of relatively low cost and high sensitivity (mainly because of their
greater thickness, ~10 mm versus only ~1 mm in ionization detectors), especially
for medium to high energy photons.
300
Side view
Detector (crystal)
Detector thickness
Detector width
Collimator length
Collimator
End view
FIG.10.6. A typical intra-operative probe (Node Seeker 900, Intra Medical Imaging LLC,
Los Angeles, CA, United States of America). (a) Hand-held detector. (b) Control and display
unit which not only displays the current count rate but also often emits an audible signal, the
tone of which is related to the count rate, somewhat analogous to the audible signal produced
by some Geiger counters. (c) A diagram of the detector and collimator assembly of a typical
intra-operative probe, illustrating that the detector (crystal) is recessed from the collimator
aperture. (Courtesy of Intra Medical Imaging LLC, Los Angeles, CA, USA.)
Disadvantages include bulkiness, and relatively poor energy resolution and

scatter rejection relative to semiconductor based probes. In some scintillation
detector intra-operative probes, the light signal from the crystal is guided to a
remote PMT through a flexible fibre-optic cable, allowing the probe assembly
to be made relatively light and compact, and more like a surgical instrument.
However, the significant loss of light in the long cable makes it more difficult to
separate scatter from unscattered X rays and rays.
On the other hand, semiconductor based probes are compact and have
excellent energy resolution and scatter rejection. To minimize structural
imperfections which degrade energy resolution, semiconductor detectors are
made relatively thin (only ~1 mm), but at the cost of lower intrinsic sensitivity.
The main disadvantage of semiconductor detectors remains their limited
thickness and resulting lower sensitivity, especially for medium to high energy
X rays and rays. Nonetheless, while scintillation detectors can be made thicker
and, therefore, more sensitive, semiconductor detectors produce more electrons
per X ray and ray stopped, and, therefore, have a superior energy resolution. To
301
CHAPTER 10
date, the few clinical studies directly comparing scintillation and semiconductor
intra-operative probes have not provided a clear choice between the two types of
probe.
10.4.5. Organ uptake probe
Historically, organ uptake probes have been used almost exclusively for
measuring thyroid uptakes and are, thus, generally known as thyroid uptake
probes.2 Thyroid uptake (i.e. the decay-corrected per cent of administered activity
in the thyroid) may be measured following oral administration of 131I-iodide,
123
I-iodide or 99mTc-pertechnetate. The uptake probe is a radionuclide counting
system consisting of a wide-aperture, diverging collimator, a NaI(Tl) crystal
(typically ~5 cm thick by ~5 cm in diameter), a PMT, a preamplifier, an amplifier,
an energy discriminator (i.e. an energy window) and a gantry (stand) (Figs 10.7(a)
and (b)). Commercially available thyroid uptake probes are generally supplied as
integrated, computerized systems with automated data acquisition and processing
capabilities, yielding results directly in terms of per cent uptake.
Each determination of the thyroid uptake includes measurement of the
thyroid (i.e. neck) count rate, the thigh background count rate (measured
over the patients thigh and presumed to approximate the count contribution of
extra-thyroidal neck activity), the standard count rate (often counted in a neck
phantom simulating the thyroid/neck anatomy) and the ambient (i.e. room)
background, with a 15 min counting interval for each measurement. Based
on the foregoing measurements, and knowing the fraction of the administered
activity which is in the standard, the thyroid uptake is calculated as follows:
uptake (%) =
C neck t neck C thigh t thigh

C standard t standard C room t room
F 100% (10.2)
where
C
t
is the total counts;

is the measurement time;
and F is the fraction of administered activity in the standard.
At one time, organ uptake probes were also used to measure kidney timeactivity data
for the evaluation of renal function. In addition, organ uptake probes have been adapted to such
well counter applications as counting of blood samples and wipes.
302
nOn-IMAGInG DEtECtORs AnD COUntERs
FIG. 10.7. (a) A typical organ (thyroid) uptake probe system, including an integrated
computer, set-up for a thyroid uptake measurement (AtomLab 950 Thyroid Uptake System,
Biodex Medical Systems, Shirley, NY, USA). The rather large neck to collimator aperture
distance (typically of the order of 30 cm) should be noted. Although this reduces the overall
sensitivity of the measurement of the neck count rate, it serves to minimize the effect of the
exact size, shape and position of the thyroid, and the distribution of radioisotope within the
gland. (b) A diagram (side view) of the open, or flat-field, diverging collimator typically used
with thyroid uptake probes. (Courtesy of Biodex Medical Systems, Inc, Shirley, NY, USA.)
by including measurement of a standard activity with each uptake

determination, corrections for radioactive decay and day to day variation in
system sensitivity are automatic. This approach is sometimes known as the
two-capsule method, since one 131i capsule is administered to the patient while
a second, identical capsule serves as the standard and is counted with each
uptake measurement. alternatively, the patient capsule itself can be measured
immediately before it is administered and then each subsequent uptake value for
radioactive decay can be corrected from the time of measurement to the time of
administration (by multiplying the right side of eq. (10.2) by et where is the
physical decay constant of the administered isotope and t is the administration
to measurement time interval). This is sometimes known as the one-capsule
method. for either the one- or two-capsule method, the fraction of administered
activity in the standard is unity. some centres administer radioiodine as a solution,
which is more cost effective, rather than as a capsule. The standard is typically
some dilution of the administered solution and the fraction of administered
activity in the standard is, therefore, an independently determined value less than
unity; for example, if the activity in the standard solution were 1/100th of the
value in the administered solution, the value would equal 0.01.
303
CHAPTER 10
Thyroid uptake measurements are now often performed by region of

interest analysis of planar scintigraphic images of the neck and of a standard
(i.e. phantom) acquired with a camera with parallel-hole collimation.
Organ uptake probes have also been used to measure total body activity,
for example, as part of individualized dosimetry based radioiodine treatment
of thyroid cancer. For this application, the patient may serve as his or her own
standard by measuring the patients total body count rate shortly (3060 min) after
administration of the radioisotope to allow it to disperse somewhat throughout
the body but before the patient has voided or otherwise excreted any of the
administered activity; in Eq. (10.3) below, this is designated time zero. Whole
body measurements are performed with the collimator removed from the probe,
the crystal oriented horizontally and at a height above the floor corresponding to
the mid-height of the patient, either seated or standing, at a distance of ~3 m from
the crystal. Further, anterior and posterior (i.e. conjugate-view) measurements are
performed by having the patient facing towards and then away from the crystal
for the respective measurements. The whole body activity (i.e. the per cent of
administered activity in the body) is then calculated based on the geometric mean
of the anterior and posterior count rates:
A B P B 1/2

t A t B t P t B
100% (10.3)
Total body activity (%) =
1/2
A(0) B(0) P(0) B(0)

t A(0) t B(0) t P(0) t B(0)
where
A and P are the anterior and posterior total body counts, respectively;
B
is the room (background) counts;
tA, tP and tB are the counting intervals for anterior, posterior and room counts,
respectively;
and (0) indicates the same quantities at time zero.
As above, the total body activity may be corrected for radioactive decay
from the time of measurement to the time of administration (by multiplying
the right side of Eq. (10.3) by et where is the physical decay constant of the
administered isotope and t is the administration to measurement time interval).
304
10.5. QUALITY CONTROL OF DETECTION AND COUNTING DEVICES

QC, which may be defined as an established set of ongoing measurements
and analyses designed to ensure that the performance of a procedure or instrument
is within a predefined acceptable range, is a critical component of routine nuclear
medicine practice. The following is a brief review of routine QC procedures for
non-imaging nuclear medicine instrumentation.
Documenting of QC procedures and organized, retrievable records of
the results of such procedures are requirements of a sound, compliant QC
programme. A written description of all QC procedures, including the acceptable
(or tolerance) range of the results of each such procedure and the corrective
action for an out of tolerance result, should be included in the facilitys procedure
manual. For each procedure, the written description should be signed and dated
by the facility director, physicist or other responsible individual. For each QC test
performed, the following data, as well as the initials or signature of the individual
performing the test, should be recorded on a structured and suitably annotated
form:
The test performed;
The date and time of the test;
The make, model and serial number of the device tested;
The make, model, serial number, activity at calibration and date of
calibration of any reference source(s) used;
The result(s) of the test;
A notation indicating whether the test result was or was not acceptable
(i.e. was or was not within the specified tolerance range).
Such records should be archived in chronological order in a secure but
reasonably accessible location. It is generally helpful to track the results of QC
tests longitudinally (e.g. in the form of a graph of the numerical result versus the
date of the test). In this way, long term trends in instrument performance, often
imperceptible from one day to the next, may become apparent. Increasingly, of
course, such records are maintained in electronic form. It is advisable, however,
to also maintain records in hard copy form, both as a backup and for convenient
review by regulators and other inspectors.
10.5.1. Reference sources
QC tests of nuclear medicine instrumentation are often performed not
with the radionuclides that are used clinically but with longer lived surrogate
radionuclides in the form of so-called reference sources. Such standards are
305
CHAPTER 10
commercially available in various activities and geometries, depending on the

application. Importantly, in the USA, the certified activities of such reference
sources must be traceable to the National Institute of Standards and Technology
(NIST), formerly the National Bureau of Standards. NIST traceability helps
ensure the accuracy of the calibrated activity. As such reference sources are
long lived, a single standard may be used for months to years, avoiding the need
to prepare sources on a daily or weekly basis and, thereby, avoiding possible
inaccuracies in dispensing activity as well as the possibility of radioactive
contamination. On the other hand, as with all sealed sources, reference sources
must be checked for leakage of radioactivity (i.e. wipe-tested) periodically and
an up to date inventory of such standards must be maintained. Reference sources
are still radioactive at the end of their useful lifespan and must, therefore, be
returned to the vendor or a third party or otherwise disposed of as radioactive
waste.
A long lived radionuclide comprising a reference source must match, in
terms of the frequency and energy of its X ray and ray emissions, the clinical
radionuclide for which it acts as a surrogate in order to ensure that instrument
responses to the clinical radionuclide and to its surrogate are comparable.
Surrogate radionuclides commonly used in nuclear medicine and their physical
properties and applications are summarized in Table 10.2.
10.5.2. Survey meter
QC tests of survey meters generally include a daily battery check, with
a display indicating whether the voltage supplied by the battery is within the
acceptable operating range. In order to confirm that the survey meter has not been
contaminated (i.e. yields a reproducibly low exposure or count rate in the absence
of radioactivity), the background exposure or count rate should be measured daily
in an area remote from radioactive sources within the nuclear medicine facility,
if such an area is reasonably accessible. In addition, survey meters should be
checked daily for constancy of response by measuring the exposure or count
rate of a long lived reference source in a reproducible measurement geometry.
Aside from the short term decay of the reference source, the measured day to
day exposure or count rates should agree within 10%; if not, the meter should be
recalibrated.
Survey meters should be calibrated that is, checked for accuracy
using suitable long lived reference sources at installation, annually and after any
repair. If the source is small (compared to the mean free path of its emitted
X rays and rays within the material comprising the source) and the distance
between the source and meter large (compared to the dimensions of the source),
306
then a point-source geometry is approximated and the expected dose rate D in

air is given by the inverse square law:
t
= A 0e
D
(10.4)
2
d
where
A0 is the activity of the reference source at calibration;
is the physical decay constant of the radionuclide comprising the reference
source;
Dt is the time interval between the calibration of the reference source and the
current measurement;
Gd is the air kerma rate constant (the subscript d indicates that only photons
with energies greater than d, typically set at 20 keV, are included) of the
radionuclide comprising the reference source;
and d is the distance between the reference source and the meter (Table 10.2).
The dose rates should be measured on each scale and, by appropriate
adjustment of the sourcemeter distance, with two readings (~20% and ~80% of
the maximum) on each scale. For all readings, the expected and measured dose
rates should agree within 10%.
Many nuclear medicine facilities have their survey meters calibrated by
the institutional radiation safety office or by a commercial calibration laboratory.
In addition to a calibration report (typically, a one page document) specifying
the reference source(s) used, the measurement procedure, and the measured
and expected exposure rates, a dated sticker summarizing the calibration results
should be affixed to the meter itself.
10.5.3. Dose calibrator
Among routine dose calibrator QC tests3, constancy must be checked
daily and accuracy and linearity at least quarterly; daily checks of accuracy
are recommended. For the constancy test, an NIST-traceable reference source,
such as 57Co, 68Ge or 137Cs (Table 10.2), is placed in the dose calibrator and the
3
At the installation of a dose calibrator, the geometry dependent response for 99mTc
must be measured and volume dependent (225 mL) correction factors relative to the standard
volume (e.g. 10 mL) derived. This procedure is required periodically following installation.
307
308
30 a
287 d
272 d
Half-life
0.0231/a
0.00241/d
0.00254/d
662 keV (86%)
511 keV (178%)
122 keV (86%)
Photopeak energy E
and frequency
of principal
X ray or ray
82.1
129
14.1
Air kerma rate

constant
(mGy m2 h1 GBq1)a
Quality control
application
Well counter
constancy and accuracy
Dose calibrator
accuracy and constancy
Well counter
Dose calibrator
Well counter
Dose calibrator
Geometry
and activity
Test tube-size rod,
~37 kBq
Vial/small bottle,
185370 MBq
Test tube-size rod,
37 kBq
Vial/small bottle,
185370 MBq
Test tube-size rod,
37 kBq
Vial/small bottle,
185370 MBq
The air kerma rate constant is equivalent to the older specific ray constant .
Germanium-68 in a sealed source is in secular equilibrium with its short lived, positron emitting daughter 68Ga (half-life: 68 min).
Cs
137
Geb
68
Co
57
Radionuclide
Physical
decay
constant
TABLE10.2. LONG LIVED RADIONUCLIDES COMPRISING REFERENCE SOURCES FOR INSTRUMENTATION

QUALITY CONTROL
CHAPTER 10
nOn-IMAGInG DEtECtORs AnD COUntERs
activity reading on each scale recorded; day to day readings should agree within
10%. for the accuracy test (sometimes also known as the energy linearity test), at
least two of the foregoing NisT-traceable reference sources are separately placed
in the dose calibrator and the activity reading on each scale recorded. for each
source, the measured activity on each scale and its current actual activity should
agree within 10%.
FIG. 10.8. Set of lead-lined plastic sleeves (CalicheckTM Dose Calibrator Linearity Test Kit,
Calicheck, Cleveland, OH, USA) for evaluation of dose calibrator linearity by the shield
method. The set is supplied with a 0.64 cm thick lead base, a colour coded unlined sleeve (to
provide an activity measurement equivalent to the zero time point measurement of the decay
method) and a six colour coded lead-lined sleeve providing attenuation factors nominally
equivalent to decay over 6, 12, 20, 30, 40 and 50 h, respectively. (Courtesy of Calicheck,
Cleveland, OH, USA.)
The quarterly check of linearity by the so-called decay method begins

with a high activity (~37 Gbq), independently calibrated 99mTc source and its
activity is assayed at 12 h intervals over three consecutive days. over that time,
equivalent to twelve half-lives of 99mTc, the activity decays to ~10 Mbq. The
measured activities are then plotted versus time on a semi-logarithmic graph and
the best fit straight line drawn through the data points thus plotted (either by
eye or using a least squares curve-fitting algorithm). for each data point, the
difference between the measured activity and the activity on the best fit straight
line at that point should be less than 10%. an alternative approach to checking
linearity is the shield method in which lead sleeves of increasing thickness
are placed in the dose calibrator with a 99mTc source (fig. 10.8). by interposing
increasing decay-equivalent thicknesses (as specified by the manufacturer for
309
CHAPTER 10
the set of lead sleeves) between the source and the dose calibrators sensitive
volume, a decay-equivalent activity is measured for each sleeve. While the shield
method is much faster than the decay method for checking linearity (taking
minutes instead of days), an initial decay based calibration of the set of sleeves
is recommended to accurately determine the actual decay equivalence of each
shield.
10.5.4. Well counter
The routine QC tests for well counters include checks of the photopeak
energy window (i.e. energy peaking) if the counter is equipped with a
multichannel analyser, background, constancy and efficiency (or sensitivity).
Prior to counting samples containing a particular radionuclide, the energy
spectrum should be checked to verify that the counter is properly peaked,
that is, that the radionuclides photopeak coincides with the preset photopeak
energy window4. For each photopeak energy window used, the background
count rate should be checked daily. Importantly, electronic noise as well as
ambient radiation levels, which may be relatively high and variable in a nuclear
medicine facility, will produce a non-zero and potentially fluctuating background
count rate. Furthermore, even trace contamination of the counting well will
produce inaccurately high count rate values. Accordingly, a blank (i.e. an
empty counting tube or vial) should always be included to determine the current
background count. To check constancy, at least one NIST-traceable reference
source (Table 10.2) should likewise be counted each day; day to day net (i.e. gross
minus background) count rates should agree within 10%.
In addition, as noted above, for each radionuclide for which a particular
well counter is used, the counter should be calibrated that is, its efficiency
(sensitivity) (in cpm/kBq) determined at installation, annually and after any
repair (Eq. (10.1)).
10.5.5. Intra-operative probe
In addition to daily battery and background checks (as done for survey
meters), QC tests of intra-operative probes should include a daily bias check
for both the primary and any backup battery to verify that bias voltage (or
high voltage) is within the acceptable range. As intra-operative probes may not
provide a display of the energy spectrum, it may not be possible to visually check
4
Isotope specific radionuclide counting or imaging with a scintillation detector is

commonly performed using a 20% photopeak energy window, equivalent to an energy range of
E 10% where E is the X ray or ray energy of the radionuclide.
310
that the probe is properly peaked, that is, that the photopeak coincides with the
preset photopeak energy window. The lower counts or count rates resulting from
an inappropriate energy window may, therefore, go unnoticed. Thus, a long lived
reference source or set of reference sources (such as 57Co, 68Ge and/or 137Cs
(Table 10.2)) should be available for daily checks of count rate constancy; a
marked change (e.g. >10%) in the net count rate from one day to the next may
indicate an inappropriate energy window setting or some other technical problem.
Ideally, the reference sources should each be incorporated into some sort of cap
that fits reproducibly over the probe so that spurious differences in count rates
due to variations in sourcedetector geometry are avoided.
10.5.6. Organ uptake probe
Aside from differences in counting geometry and sensitivity, uptake probes
and well counters actually have very much in common and the QC procedures
checks of the photopeak energy window, background, constancy and
efficiency are, therefore, analogous. Importantly, however, efficiency should
be measured more frequently for each patient than for a well counter, so
that the probe net count rates can be reliably converted to thyroid uptakes for
individual patients.
BIBLIOGRAPHY
CHERRY, S.R., SORRENSON, J.A., PHELPS, M.E., Physics in Nuclear Medicine, 3rd edn,
NINKOVIC, M.M., RAICEVIC, J.J., ANDROVIC, A., Air kerma rate constants for emitters
used most often in practice, Radiat. Prot. Dosimetry 115 (2005) 247250.
ZANZONICO, P., Routine quality control of clinical nuclear medicine instrumentation: A brief
review, J. Nucl. Med. 49 (2008) 11141131.
ZANZONICO, P., HELLER, S., Physics, instrumentation, and radiation protection, Clinical
Nuclear Medicine (BIERSACK, H.J., FREEMAN, L.M., Eds), Springer Verlag, Berlin
Heidelberg (2007) 133.
311
CHAPTER 11
NUCLEAR MEDICINE IMAGING DEVICES
M.A. LODGE, E.C. FREY
Russell H. Morgan Department of Radiology and Radiological Sciences,
Johns Hopkins University,
Baltimore, Maryland, United States of America
11.1. INTRODUCTION
Imaging forms an important part of nuclear medicine and a number of
different imaging devices have been developed. This chapter describes the
principles and technological characteristics of the main imaging devices used
in nuclear medicine. The two major categories are gamma camera systems and
positron emission tomography (PET) systems. The former are used to image
rays emitted by any nuclide, while the latter exploit the directional correlation
between annihilation photons emitted by positron decay. The first section of this
chapter discusses the principal components of gamma cameras and how they
are used to form 2D planar images as well as 3D tomographic images (single
photon emission computed tomography (SPECT)). The second section describes
related instrumentation that has been optimized for PET data acquisition. A major
advance in nuclear medicine was achieved with the introduction of multi-modality
imaging systems including SPECT/computed tomography (CT) and PET/CT. In
these systems, the CT images can be used to provide an anatomical context for
the functional nuclear medicine images and allow for attenuation compensation.
The third section in this chapter provides a discussion of the principles of these
devices.
11.2. GAMMA CAMERA SYSTEMS
The gamma camera, or Anger camera [11.1], is the traditional workhorse of
nuclear medicine imaging and its components are illustrated in Fig.11.1. Gamma
camera systems are comprised of four basic elements: the collimator, which
defines the lines of response (LORs); the radiation detector, which counts incident
photons; the computer system, which uses data from the detector to create 2D
312
rays
Collimator
Aluminium cover
Scintillation crystal
Reflector
Exit window
Photomultiplier tube array
Preamplifiers
Shaping and positioning
Corrections and framing
FIG.11.1. Schematic diagram showing the major components of a gamma camera.
histogram images of the number of counted photons; and the gantry system, which
supports and moves the gamma camera and patient. The overall function of the
system is to provide a projection image of the radioactivity distribution in the
patient by forming an image of rays exiting the body. Forming an image means
establishing a relationship between points on the image plane and positions in the
object. This is sometimes referred to as an LOR: ideally, each position in the
image provides information about the activity on a unique line through the object.
In gamma cameras, single photons are imaged, in contrast to PET where pairs of
photons are imaged in coincidence. Thus, in order to define an LOR, a lens is
required, just as in optical imaging. However, the energies of rays are so high
Profile through image
Image plane
Collimator
Point sources
FIG.11.2. The drawing on the left demonstrates the image of two point sources that would
result without the collimator. It provides very little information about the origin of the photons
and, thus, no information about the activity distribution in the patient. The drawing on the
right illustrates the role of the collimator and how it defines lines of response through the
patient. Points on the image plane are uniquely identified with a line in space.
313
CHAPTER 11
that bending them is, for all practical purposes, impossible. Instead, collimators
are used to act as a mechanical lens. The function of a collimator is, thus, to define
LORs through the object. Figure11.2 illustrates the function of and need for the
collimator, and, thus, the basic principle of single photon imaging.
11.2.2. The Anger camera
11.2.2.1. Collimators
As mentioned above, the collimator functions as a mechanical lens: it
defines LORs. The collimator accomplishes this by preventing photons emitted
along directions that do not lie along the LOR from reaching the detector. Thus,
collimators consist of a set of holes in a dense material with a high atomic
number, typically lead. The holes are parallel to the LORs. Ideally, each point in
the object would contribute to only one LOR. This requires the use of collimator
holes that are very long and narrow. However, such holes would allow very few
photons to pass through the collimator and be detected. Conversely, increasing
the diameter or decreasing the length of the holes results in a much larger range
of incident angles passing through the collimator. As illustrated in Fig.11.3, this
results in degraded resolution. As can be seen from this figure, each hole has a
cone of response and the diameter of the cone of response is proportional to the
distance from the face of the collimator.
As discussed above, changing the diameter of collimator holes changes
the resolution and also the number of photons passing through the collimator.
The noise in nuclear medicine images results from statistical variations in the
number of photons counted in a given counting interval due to the random nature
of radiation decay and interactions with the patient and camera. The noise is
described by Poisson statistics, and the coefficient of variation (per cent noise)
is inversely proportional to the square root of the number of counts. Thus,
increasing the number of counts results in less noisy images. As a result, there
is an inverse relationship between noise and spatial resolution for collimators:
improving the resolution results in increased image noise and vice versa.
Another important characteristic of collimators is the opacity of collimator
septa to incident rays. In an ideal collimator, the septa would block all incident
radiation. However, in real collimators, some fraction of the radiation passes
through or scatters in the septa and is detected. These phenomena are referred
to as septal penetration and scatter. The amount of septal penetration and scatter
depends on the energy of the incident photon, the thickness and composition of
the septa, and the aspect ratio of the collimator holes. Since gamma cameras are
used to image radionuclides with energies over a wide range, collimators are
typically available that are appropriate for several energy ranges: low energy
314
collimators are designed for isotopes emitting photons with energies lower than
approximately 160keV; medium energy collimators are designed for energies up
to approximately 250keV; and high energy collimators are designed for higher
energies. It should be noted that in selecting the appropriate collimator for an
isotope, it is important to consider not only the photon energies included in the
image, but also higher energy photons that may not be included in the image.
For example, in 123I there are a number of low abundance high energy photons
that can penetrate through or scatter in septa and corrupt the images. As a result,
medium energy collimators are sometimes used for 123I imaging, despite the main
photopeak at 159keV.
Point sources
Collimator
holes
FIG.11.3. Illustration of the concept of spatial resolution and how collimator hole length
and diameter affect spatial resolution. The lines from the point source through the collimator
indicate the furthest apart that two sources could be and still have photons detected at the
same point on the image plane (assumed to be the back of the collimator). Thus, sources closer
together than this would not be fully resolved (though they might be partially resolved). From
this, we see that the resolution decreases as a function of distance. It should also be noted that
the resolution improves proportionally with a reduction in the width of the collimator holes
and improves (though not proportionally) with the hole length.
For multi-hole collimators, hole shape is an additional important factor in

collimator design. The three most common hole shapes are shown in Fig.11.4.
Round holes have the advantage that the resolution is uniform in all directions
in planes parallel to the face of the collimator. However, as discussed below, the
315
CHAPTER 11
sensitivity is relatively low because there is less open area for a given resolution
and septal thickness. Hexagonal hole collimators are the most common design for
gamma cameras using continuous crystals. They have the advantage of relatively
direction independent response functions and higher sensitivity than a round hole
collimator with the same resolution and septal thickness. Square hole collimators
are especially appropriate for detectors that have pixelated crystals. Having
squares holes that match the spacing and inter-gap distance of these detectors
results in good sensitivity with these detectors. However, the resolution varies
significantly depending on the direction, being worse by a factor of a 2 along
the diagonal direction compared to parallel to the rows of holes.
FIG.11.4. Examples of the three major hole shapes used in multi-hole collimators. They are,
from left to right: round, hexagonal and square holes. In all cases, black indicates septa and
white open space. The diameter (often called flat-to-flat for square and hexagonal holes) is d
and the septal thickness is s.
Multi-hole collimators typically have many thousands of holes. The

uniformity of the image critically depends on the holes having uniform sizes and
spacing. As a result, high quality fabrication is essential. The septa must be made
of a high density, high Z material in order to stop the incident rays. Lead is
the material of choice for most multi-hole collimators due to its relatively low
cost, low melting temperature and malleability. Lead multi-hole collimators
can be divided into two classes based on fabrication techniques: cast and foil
collimators.
Fabrication of cast collimators involves the use of a mould that is filled with
molten lead to form the collimator. The mould typically consists of two plates
with holes at opposing positions that match each of the holes in the collimator. A
set of pins is placed in the holes. Lead is then poured between the plates and the
entire assembly is then carefully cooled. The plates and pins are removed leaving
behind the collimator. This technology is especially well suited to making high
and medium energy collimators as well as special purpose collimator geometries.
316
Foil collimators are created from thin lead foils. The foils are stamped and
then glued together to build up the collimator layer by layer. Figure11.5 shows a
schematic of how two layers are stamped and glued to form the holes. It should be
noted that in the stamping the septa that are glued must be thinner than the other
walls in order to retain uniform septal thickness and, thus, maximize sensitivity.
It is clear that precise stamping, alignment and gluing is essential to form a high
quality collimator. The septa in foil collimators can be made thinner than in cast
collimators. As a result, foil fabrication techniques are especially appropriate
for low energy collimators. Understanding the fabrication technology can help
in diagnosing problems with the collimator. For example, Fig.11.6 shows an
image with non-uniformities appearing as vertical stripes in the image of a sheet
source. This was a foil collimator and the non-uniformities apparently originated
from fabrication problems, resulting in some layers having different sensitivities
compared to other layers.
FIG.11.5. Illustration of fabrication of foil collimator by gluing two stamped lead foils. It
should be noted that the foils must be stamped so that the portions of the septa that are glued
are half the thickness of the rest of the septa. Furthermore, careful alignment is essential to
preserve the hole shapes.
FIG.11.6. Uniformity image of a defective foil collimator. The vertical stripes in the image
result from non-uniform sensitivity of the collimator due to problems in the manufacturing
process. The peppery texture is due to quantum noise and is visible because the grey level was
expanded to demonstrate the non-uniformity artefacts.
317
CHAPTER 11
Focal point
Lines of response
Lines of response
Imaging plane
Lines of response
Pinhole
Lines of response
Imaging plane
Imaging plane
Imaging plane
Focal point
FIG.11.7. Illustration of the four common collimator geometries: (left to right) parallel,
converging, diverging and pinhole.
A final important characteristic of collimators is the hole geometry. There

are four common geometries in nuclear medicine: parallel, converging, diverging
and pinhole, as illustrated in Fig.11.7. Parallel holes are the most commonly used
collimators. The LORs are parallel, and there is, thus, a one to one relationship
between the size of the object and image. In converging hole collimators, the
LORs converge to a point (focal point) or line (focal line) in front of the collimator,
and there is, thus, magnification of the image. These two collimator types are
referred to as cone-beam and fan-beam collimators, respectively. These are useful
for imaging small objects, such as the heart or brain, on a large camera as they
provide an improved trade-off between spatial resolution and noise. In diverging
hole collimators, the LORs converge to a point or line behind the collimator. This
results inminification of the image. Diverging hole collimators are useful for
imaging large objects on a small camera. However, they result in a poor resolution
versus noise trade-off, and are, thus, infrequently used. Pinhole collimators
use a single hole to define the LORs. In terms of geometry, they are similar to
cone-beam collimators but with the focal point between the image plane and the
object being imaged. As a result of this, the image is inverted compared to the
object. In addition, the object can be eitherminified or magnified depending on
whether the distance from the image plane to the focal point is less than or greater
than the distance from the pinhole to the object plane. Pinhole collimators provide
an improved resolution noise trade-off when objects are close to the pinhole. They
are, thus, useful for imaging small objects such as the thyroid or small animals.
Another advantage of pinholes is that there is only a single hole, referred to as an
aperture, which determines the amount of collimator penetration and scatter. As
a result, it is possible to fabricate the aperture from high density and high atomic
number materials (e.g. tungsten, gold or depleted uranium), which can reduce
318
collimator penetration and scatter. This makes these collimators appropriate for
imaging radionuclides emitting high energy rays such as 131I. In addition, pinhole
collimators with changeable apertures can have different diameter pinholes. This
allows selection of resolution/sensitivity parameters relatively easily.
The collimator properties can be most completely described by the collimator
point source response function (PSRF), the noise-free image of a point source in
air with unit activity using an ideal radiation detector, as a function of position
in the object space. The shape of the collimator PSRF completely describes the
resolution properties, and when normalized to unit total volume is referred to as
the collimator point spread function (PSF).
Figure11.8 shows some sample collimator PSFs for an 131I point source
imaged with a high energy general purpose collimator and a medium energy
general purpose collimator. These PSFs are averaged over the position of the
source with respect to the collimator and, thus, do not show the hole pattern. There
are several things to note from this figure. First, using a properly designed
collimator reduces septal scatter and penetration to very low levels, while they
become significant for a collimator not designed to handle the high energies of
131
I. Second, the response function becomes wider as a function of distance,
demonstrating the loss of resolution as a function of increasing source to
collimator distance. Finally, there is evidence of the shape of the holes, which, in
this case, were hexagonal. The shape can be barely discerned in the shape of the
central portion of the response, which is due to photons passing through the
collimator holes. The geometry of the collimator is more evident in the septal
FIG.11.8. Sample images of the point spread function for an 131I point source at (left to right)
5, 10 and 20 cm from the face of a high energy general purpose collimator (top row) and
a medium energy general purpose collimator (bottom row). The images are displayed on a
logarithmic grey scale to emphasize the long tails of the point spread function resulting from
septal penetration and scatter. The brightness of the image has been increased to emphasize
the septal penetration and scatter artefacts.
319
CHAPTER 11
penetration and scatter artefacts. In fact, the septal penetration is highest along
angular directions where the path through the septa is thinnest, giving rise to the
spoke-like artefacts in the directions perpendicular to the walls of the hexagonal
holes.
Another useful way to describe and understand the resolution properties of
the collimator is in terms of its frequency response. This can be described by the
collimator modulation transfer function, which is the magnitude of the Fourier
transform of the collimator PSF. Figure11.9 shows some sample profiles through
the collimator modulation transfer function. It should be noted that the collimator
response does not pass high frequencies very well and, for some frequencies,
the response is zero. This attenuation of high frequencies results in a loss of
fine detail (i.e. spatial resolution) in the images. Finally, the cut-off frequency
decreases with distance from the collimator and different collimator designs have
different frequency responses.
FIG.11.9. Sample profile through the geometric modulation transfer functions (MTFs) for
low, medium and high energy (HE, ME and LE, respectively) general purpose (GP) and high
resolution (HR) collimators for a source 5 cm (left) and 20 cm (right) from the face of the
collimator.
It is often desirable to summarize the collimator resolution in terms of one

or two numbers rather than the entire response function. This is often done in
terms of the width of the collimator PSRF at a certain fraction of its maximum
value. For example, Fig.11.10 shows a sample profile through a collimator
PSF and the position of the full width at half maximum (FWHM) and the full
width at tenth maximum (FWTM). To good approximation, the FWHM of a
collimator is proportional to the distance from the face of the collimator. This
holds for all distances except those very close to the collimator face, as illustrated
in Fig.11.11. The FWTM is useful for assessing the amount of septal penetration
and scatter that are present. For a Gaussian response function, which is a good
320
FWHM
Intensity (arbitrary units)
1.0
0.8
0.6
Half maximum
0.4
0.2
0.0
FWTM
-2
-1.5
-1
Tenth maximum
-0.5 0
0.5
Distance (cm)
1.5
FIG.11.10. Plot of the total collimatordetector point spread function for a medium energy
general purpose collimator imaging 131I, indicating the positions of the full width at half
maximum (FWHM) and the full width at tenth maximum (FWTM).
2.5
HEGP
MEGP
LEGP
LEHR
FWHM (cm)
2.0
1.5
1.0
0.5
0.0
10
15
20
25
30
35
Distance from face of collimator (cm)

FIG.11.11. Plot of the full width at half maximum (FWHM) of the geometric collimator
detector point source response function including a Gaussian intrinsic response with an
FWHM of 4mm as a function of distance from the face of the collimator for the same set of
collimators described in Fig.11.9.
321
CHAPTER 11
approximation for the combination of a Gaussian intrinsic detector response

with the geometric collimator response, the relationship between the FWHM and
FWTM is given by FWTM/FWHM 1.86. Thus, if the FWTM is substantially
larger than this factor times the FWHM, the response has been affected by factors
other than the geometric response, such as septal penetration and scatter. It should
be noted that the effects of septal penetration and scatter on the FWTM are less
visible in a PSRF than they are in a line source response function.
The FWHM of the collimator resolution can be estimated from geometric
arguments. Figure11.12 shows a schematic that can be used to derive the
resolution for a point source a distance Z from the face of the collimator. The
collimator hole has a length L and a width d. The image plane (often taken to be
the mean path of the primary photons in the crystal plus any physical gap) is a
distance B behind the back face of the collimator. The photon passing through the
collimator holes with the most oblique angle of incidence will have an incident
angle defined by tan q=d/L. Thus, the extreme limits of the response function
will be defined by this limit. If it is assumed that the geometric response function
is triangular in shape, then the FWHM in Fig.11.12 will be half of this distance.
Using similar triangles, it can be shown that the FWHM is given by:
FWHM =
d
(Z + L + B) (11.1)
L
Thus, it can be seen that, as described above, the FWHM is linearly related
to the distance from the face of the collimator and is proportional to that distance
when the distance is large compared to L + B.
The resolution of the collimatordetector system depends on both the
resolution of the collimator and the intrinsic resolution of the gamma camera. For
continuous-crystal gamma cameras, the intrinsic resolution can be modelled with
a Gaussian function. In this case, the total response function for the collimator
detector is the convolution of the intrinsic resolution and the collimator response.
If the collimator response is approximated by a Gaussian function, the FWHM is
given by the Pythagorean sum of the intrinsic and collimator FWHMs:
2
2
FWHM total = FWHM collimator
+ FWHM intrinsic
(11.2)
Figure11.11 shows a plot of the total geometric FWHM resolution

for several collimators including the effects of a 4mm Gaussian intrinsic
resolution. The curves through the points represent a fit with a function of the
above Pythagorean sum. It should be noted that except for distances close to
the collimator, the resolution is linear with distance, indicating that the total
resolution is dominated by the collimator resolution.
322
Collimator
Point
source
Image plane
FIG.11.12. Diagram illustrating the collimator geometry used to derive the expression for the
full width at half maximum.
The integral of the collimator PSRF gives the sensitivity of the collimator.
This is, in principle, a dimensionless quantity which gives the fraction of emitted
photons that pass through the collimator, and is of the order of 103104 for
typical nuclear medicine collimators. It is often useful to express the sensitivity
in terms of counts per unit activity per unit time, for example counts per second
per megabecquerel. For a parallel-hole collimator, the sensitivity is a function
of two terms: the solid angle of the hole, which is a function of (d/L)2, and the
fraction of the active area of the collimator that is open area (hole) as compared
to septa. The second term can be described in terms of the unit cell: the smallest
geometric region that can be used to form the entire collimator by a set of simple
translations. The sensitivity S of a parallel-hole collimator is given by:
S=
aopen aopen
4 L2 a total
(11.3)
where
aopen is the open area of a collimator unit cell, i.e. the area of the hole itself;
and atotal is the total area of the cell including the part of the collimator septa lying
in the unit cell.
323
CHAPTER 11
From the above it can be seen that, for a parallel-hole collimator, the
sensitivity is independent of the distance to the collimator face. This is because
there is a balance between the decrease in sensitivity from a single hole and the
increase in the number of holes through which photons can pass as a function
of increasing distance from the collimator face. It should also be noted that
aopen is proportional to d2, which is proportional to the FWHM resolution. The
term aopen/atotal varies slowly as a function of d if d s. Thus, the sensitivity is
proportional to the square of the resolution:
S = k FWHM 2 (11.4)
where the constant k depends only weakly on the FWHM. Since noise is directly
related to the number of counts, there is a fundamental trade-off between
resolution and noise. From the above, it is also evident that maximizing the ratio
of aopen/atotal is important in terms of reducing noise for a given resolution.
11.2.2.2. Scintillation crystals
The scintillation crystal in the gamma camera converts ray photons
incident on the crystal into a number of visible light photons. The characteristics
and principle of scintillation radiation sensors are described in more detail in
Chapter6. Ideally, the crystal would be dense and of a high Z material in order
to stop all incoming rays with photoelectric events. It should have high light
output to provide low quantum noise for energy and position estimation. The
decay time of the light output needs to be fast enough to avoid a pile-up of
pulses at the count rates experienced in nuclear medicine imaging procedures.
The wavelength spectrum of the scintillation photons should be matched to the
sensitivity of the photodetectors used to convert the scintillation signal into an
electrical signal. In addition to these technical properties that directly affect
image quality, there are a number of desirable material properties that influence
the cost of the device. These include the cost of the raw material, the ease of
growing large single crystals and the sensitivity to environmental factors such as
humidity. Owing to its unique combination of desirable properties, the crystals
used in gamma cameras based on photomultiplier tubes (PMTs) are typically
made of NaI(Tl). Gamma cameras based on solid state photodetectors require
a different light spectrum and typically CsI(Tl) is used in these devices. The
scintillation properties of these materials are discussed in detail in Chapter6.
As will be described below, the interaction position of the ray with the
detector is estimated based on the distribution of the scintillation light to an
array of PMTs. It is important that the distribution of light be as independent as
possible of the depth of interaction in the crystal and depends in a predictable
324
way on the lateral position. Further, absorption of scintillation photons by

defects in the crystal is highly undesirable, as it will adversely affect the
accuracy and precision of energy and position estimation. Thus, in order to make
the lateral light distribution predictable, as even as possible and tominimize
absorption, Anger cameras employ a large single crystal equal to the size of the
field of view (FOV) of the camera. This can be as large as 60cm40cm in
modern cameras. As NaI(Tl) is hygroscopic, it is important to hermetically seal
the crystal in an enclosure. The back of the crystal must be optically coupled
to the photodetector, so that the back part of the crystal enclosure consists of
an optical glass exit window optically coupled to the crystal. The exit window
lies between the crystal and the photodetector array. It serves several functions.
First, it serves to hermetically seal the crystal. Second, the exit window allows
scintillation photons to pass from the crystal into the photodetector array, and,
thus, must be transparent in the emission range of the scintillator. To reduce
internal reflections, it is desirable that the index of refraction be matched
as closely as possible to that of the scintillator (n=1.85 for NaI(Tl)) and the
photodetector (n 1.5 for borosilicate glasses used in the entrance windows
of PMTs). The remainder of the enclosure should be light-tight to block out
ambient light. The front face should be thin and of a low Z material typically
Al is used to reduce the probability of incident ray absorption. Finally, to
help collect incident light photons and improve energy resolution, the inside of
the enclosure is a reflective layer. The use of specular versus diffusive reflectors
affects the nature of the light response and has an impact on the variation in the
light response on the photodetectors as a function of interaction depth, and, thus,
impacts the precision of the position estimation.
One important parameter of the scintillation crystal related to camera
performance is its thickness. The thickness is a trade-off between two
characteristics: intrinsic resolution and sensitivity. The intrinsic resolution
depends on the crystal thickness via the variation in the light distribution as a
function of depth of interaction. Since the depth of interaction can vary over
a wider range in a thicker crystal, there will be a larger variation in the light
distribution and, thus, a larger uncertainty in the estimated lateral position of
the interaction. In other words, thicker crystals generally have poorer intrinsic
resolution. The functional relationship between the thickness and intrinsic
resolution is complicated and depends on the details of the surface treatment of
the scintillator, the photodetector array and the position estimation algorithm.
For GE Millenium VG cameras, the FWHM intrinsic resolution for 140keV
photons using 0.953, 1.587 and 2.54cm thick crystals is 3.5, 3.9 and 5.2mm,
respectively.
325
CHAPTER 11
The intrinsic sensitivity of the crystal is related to crystal thickness by:

S i = 1 e t (11.5)
where
Si

is the intrinsic sensitivity;

is the linear attenuation coefficient of the crystal;
and t is the crystal thickness.

Since the linear attenuation coefficient decreases with energy, the intrinsic
sensitivity also decreases with energy. Figure11.13 shows a plot of the intrinsic
sensitivity as a function of energy for several crystal thicknesses. For 140keV, the
sensitivity is ~92% for a 0.953cm (3/8 in) thick crystal. This is the most common
crystal thickness in commercial systems, though cameras with 5/8 and 1 in are
available. These crystal thicknesses provide substantially improved sensitivity
for radionuclides emitting medium and high energy photons such as 111In and 131I
at the cost of a relativelyminor reduction in intrinsic spatial resolution.
Intrinsic sensitivity
Crystal thickness
FIG.11.13. Plot of the intrinsic sensitivity of a NaI scintillation crystal as a function of energy
for several crystal thicknesses.
326
A final important property of the scintillation crystal is the light output. This
is a characteristic of the scintillator material, and is the number of scintillation
photons per unit energy deposited in the crystal by a photon. Thus, the total light
is proportional to the energy deposited in the crystal, and can, therefore, be used
to estimate the energy of the ray. The number of scintillation photons produced
for a given event is a Poisson random variable. Thus, the larger the number of
scintillation photons the smaller the coefficient of variation (standard deviation
divided by the mean) of the mean number of photons, and, hence, the estimated
photon energy. Thus, scintillators with high light output will provide higher
energy resolution. In addition, as will be seen below, the light distribution over
the photodetector array is used to estimate the interaction position. Since the light
collected by each element in the array is also a Poisson random variable that is
proportional to the light output, a larger light output will result in higher precision
in the estimated position, and, thus, improved intrinsic spatial resolution. One
reason that NaI(Tl) is used in gamma cameras is its high light output.
11.2.2.3. Photodetector array
The next element in the radiation detector is the photodetector array.
This array measures the distribution of scintillation photons incident on the
array and converts it into a set of pulses whose charge is proportional to the
number of scintillation photons incident on each corresponding element in
the array. As described below, the output of this array is used to compute the
interaction position of the ray in the scintillator. In clinical gamma cameras, the
photodetector array is comprised of a set of 3090 PMTs arranged in a hexagonal
close packed arrangement, as illustrated in Fig.11.14. More details on the
operation and characteristics of PMTs are provided in Chapter6. In brief, PMTs
have the advantage that they are very well understood, have a moderate cost, are
relatively sensitive to low levels of scintillation light and have a very high gain. In
some commercial designs, PMTs have been replaced by semiconductor detectors
such as photodiodes. Generally, these devices are somewhat less sensitive and
have a lower gain than PMTs, resulting in more noise in the charge signal and,
thus, less precision in the energy and position estimated from the charge signal.
Since the position and energy are estimated from the set of charge signals
from the elements in the photodetector array, it is highly desirable that the
proportionality constants relating light intensity to charge be the same for all
of the photodetectors. This can be ensured by choosing matching devices and
by carefully controlling and matching the electronic gain. For PMTs, the gain is
controlled by the bias voltage applied to the tubes. Since gain is also a function of
temperature, the temperature of the photodetectors must be carefully controlled.
The gains of PMTs are very sensitive to magnetic fields, even those as small
327
CHAPTER 11
as the Earths magnetic field. Thus, the PMTs must be magnetically shielded
using mu-metal. Finally, since the gains of tubes can drift over time, periodic
recalibration is necessary.
One of the major advantages of the gamma camera is that the number of
PMTs is much smaller than the number of pixels in images from the gamma
camera. In other words, in contrast to semiconductor detectors where a separate
set of electronics is required for each pixel, the gamma camera achieves a great
reduction in cost and complexity by estimating the interaction position of the
ray based on the output of the array of PMTs.
To understand the position estimation process, Fig.11.15 is considered.
This figure shows a cross-section through two PMTs, and the crystal and exit
window. The number of photons collected by a PMT directly (i.e. without
reflection) will be proportional to the solid angle subtended by it at the interaction
point. As can be seen in the figure, the interaction position is offset to the right,
and there is a smaller solid angle subtended by PMT 1 than by PMT 2. Thus, the
signal from PMT 1 will be smaller than for PMT 2. If the interaction position
moves to the left, so that it lies along the line separating the two PMTs, there will
be an equal amount of light collected by each PMT. The relationship between the
light collected by the two PMTs and the lateral interaction position can be used
to estimate the interaction position, as will be described in more detail below. In
addition, the total scintillation light collected by all of the PMTs is proportional
to the energy deposited by the ray in the crystal. Thus, the total charge can be
used to estimate the energy of the photon.
FIG.11.14. Cross-section of a gamma camera at the back face of the entrance window
showing the hexagonal close packed array of photomultiplier tubes. The dotted line indicates
the approximate region where useful images can be obtained.
328
Exit window
Crystal
Interaction position
FIG.11.15. Cross-section through two photomultiplier tubes (PMTs), the exit window and
crystal in a gamma camera. The interaction position of a ray photon is indicated. The solid
angles subtended by PMT 1 and 2 are 1 and 2, respectively.
11.2.2.4. Amplifiers and pulse shaping

The charge pulse from each PMT is very small and, thus, subject to
noise. In addition, the scintillation photons are emitted randomly over a finite
time (given by the scintillators decay constant), making the charge pulse rather
noisy. To make subsequent analysis of the pulse easier and more resistant to
electrical noise, the pulse is amplified and shaped prior to processing to estimate
the interaction position and photon energy. The components of this stage are a
preamplifier and shaping amplifier.
The preamplifier integrates the charge pulse from the PMT to form a
voltage pulse with height proportional to the input charge. The design of the
preamplifier should be such that the voltage height is as independent as possible
of the details of the charge pulse, such as decay and rise times. Preamplifiers are
typically mounted directly to the PMT outputs in order to avoid corruption of the
tiny charge pulses by electrical noise and interference.
Ideally, output pulses would have a very flat top to allow easy digitization
of the pulse height and be very narrow to allow high pulse rates without pulse
pile-up. However, the output pulse from the preamplifier typically has a relatively
long decay time and is not very suitable for digitization and handling high pulse
rates. As a result, the output of the preamplifier is fed into a shaping amplifier.
Typically, shaping amplifiers use a combination of integration and differentiation
stages to produce near Gaussian pulses. It should be noted that more recent
commercial gamma cameras have used digital pulse processing methods to
329
CHAPTER 11
perform this function. This involves digitizing the output waveform from the
preamplifier. This has a number of advantages including providing the ability to
change the trade-off between energy resolution and count rate, depending on the
requirements of the particular imaging procedure. In addition, this method also
provides digital estimates of the pulse heights that can be used in digital position
and energy estimation algorithms.
11.2.2.5. Position and energy estimation
The goal of the radiation detector is to provide an estimate of the energy
and interaction position of each ray incident on the detector. The output of
the photodetector array and amplifier system is a set of voltage signals for each
photon. The sum of these voltages is proportional to the gamma camera energy
and the position is a function of the set of voltage values. The position and energy
estimation circuits estimate the ray energy and position from the set of voltage
values from the photodetector array.
One way of doing this is to use a resistive network to divide the signals
from the array elements among a set of four signals often referred to as X+, X, Y+
and Y, as illustrated in Fig.11.16. The resistor values for each PMT are chosen
so that the charge is divided in proportion to its position with respect to the centre
of the array. For example, for a PMT in the centre horizontally, the resistances
for X+ and X would be equal. Similarly, for a PMT in the centre vertically, the
resistances for Y+ and Y would be equal.
Using the scheme described above, the energy E can be computed using:
E = X + + X + Y+ + Y (11.6)
However, one limitation of this method is that the total amount of light
collected is dependent on position. For example, if the interaction is directly
under a PMT, a larger fraction of the total light will be collected, resulting in a
larger value of E than if the interaction is in the gap between PMTs. This means
that the estimate of the energy will vary spatially. As discussed below, this has an
impact on camera uniformity. As a result, the energy must be corrected based on
the interaction position.
Under the assumption that the light collected by a PMT is proportional to
the distance from its centre, and with the correct resistor values, the interaction
position, defined by x and y can be computed using:
x=
330
X+ X
E
and y =
Y+ Y
(11.7)
E
In early gamma cameras, the computations above were performed using

analogue circuits. However, in recent cameras, the pulse heights (or the entire
pulse) are digitized and the computations performed digitally.
Using the resistive summing and simple estimation approaches above
results in a number of problems. First, light collected by phototubes is not linearly
related to the distance from the interaction point. For example, the amount of light
changes relatively little for PMTs at a large lateral distance from the interaction
position. As a result, thresholding circuits are often added to exclude the signal
from PMTs with small outputs (and, thus, far from the interaction point) from
the position calculation. In addition, the distribution of light between two tubes
changes more quickly when the interaction position lies between two tubes than
it does when the interaction position is directly over a tube. This results in spatial
non-linearities where images of line sources are bent towards the centre of PMTs.
A final difficulty is that it is not possible to reliably estimate the position of
photons interacting near the edge of the camera. In this case, almost all of the
light will be collected by the nearest PMT and there will be little change in the
relative amount of light as the interaction position moves.
FIG.11.16. Illustration of a resistive network used to implement position estimation. The

output from each photomultiplier tube/preamplifier is divided by a resistive network with four
outputs, X+, X, Y+ and Y. The corresponding signals from all of the photomultiplier tubes are
connected to provide the summed signal.
331
CHAPTER 11
As a result of the above difficulties, modern cameras use sophisticated

correction and position estimation methods. The correction methods will be
discussed in more detail below. Advanced position estimation methods involve
digitizing the outputs of all of the PMTs and using them in the position estimation.
In this case, position and energy estimation, and the various corrections are
done by a digital signal processor or microprocessor, allowing a great deal more
sophistication in the choice of algorithms. In some systems, these are done using
statistical estimation techniques such as maximum-likelihood estimation.
As alluded to above, the values of the interaction position (x, y) and energy
E are computed using equations similar to those shown above. The input variables
X+, X, Y+ and Y are related to the charge signal from the PMT. These signals
are proportional to the number of photoelectrons emitted from the photocathode.
The emission of photoelectrons is the end result of a series of random processes
that includes the number of scintillation photons produced, the number of these
collected by the PMT, the number of photoelectrons produced for a given photon
and the number of photoelectrons emitted which are focused onto the first
dynode. The net result is that there is statistical variation in the values of the
input variables for a given interaction position and energy. Thus, the values of
position and energy are not exact, and are only estimates of the true quantities.
Thus, there will be imprecision in the energy and position estimates resulting in
finite intrinsic energy and spatial resolutions (seeChapter8 for more details).
To a good approximation, both the energy and intrinsic spatial resolution
can be characterized by a Gaussian function. The energy resolution results from
variations in the total number of photoelectrons incident on the first dynode. The
random variations can be approximated by a Poisson distribution and the variance
in the energy resolution is, thus, approximately equal to the number of these
photoelectrons. The approximate energy resolution of a gamma camera can, thus,
be estimated as follows. A NaI(Tl) crystal produces, on average, 38 photons/keV.
The quantum efficiency (fraction of incident scintillation photons that produce
photoelectrons) of a PMT for the 415 nm emission of NaI(Tl) is approximately
12%. Thus, for a 140keV photon, the number of photoelectrons collected
is 140380.12=638 electrons. The FWHM is equal to approximately
2.35 times the standard deviation, so the fractional energy resolution is equal to
2.35 638 / 638 = 9.3%. This is approximately equal to the energy resolution for
a state of the art scintillation camera. It should also be noted from the above that
the energy resolution is proportional to E0.5 and spatial variations in the collection
efficiency will produce spatial variations in the energy resolution. Estimating the
intrinsic spatial resolution is more difficult than estimating the energy resolution
because of the more complicated estimation equation. However, typical intrinsic
spatial resolutions are in the range of 35mm, depending on the number of PMTs
used and details of the estimation procedure.
332
11.2.2.6. Corrections
As mentioned above, the energy and position estimation are non-ideal,
resulting in errors in energy and position estimates. These errors give rise to
non-uniform sensitivity in the camera. Thus, to obtain clinically acceptable
images, energy, spatial and uniformity corrections are needed. The need for
these corrections is illustrated in Fig.11.17. An image is shown resulting from
a uniform distribution of rays on the camera with the collimator removed,
often referred to as an intrinsic flood image. The substantial non-uniformity, the
presence of edge packing artefacts near the edge of the FOV, and the visibility of
the tube pattern should be noted.
Edge packing artefact
Tube pattern
FIG.11.17. Intrinsic flood image of gamma camera without energy, spatial or sensitivity
corrections.
Energy corrections are needed because the estimated energy depends on

spatial position. This behaviour can be understood in terms of variations of the
fraction of the scintillation light collected as a function of interaction position.
Since the energy is proportional to the light collected, differences in the fraction
of light collected will result in a proportional change in the estimated energy.
Figure11.18 shows an example energy spectrum where the fraction of collected
light was 2% lower or 2% higher than the average. This results in energy spectra
333
CHAPTER 11
shifted to lower or higher energy, respectively. Only photons falling within

the acquisition energy window, in this case having a full width of 20% of the
photopeak energy, centred at 140keV, are counted. There are about 1.7% fewer
photons counted in the two sample energy spectra than for the case of the average
energy spectrum. Thus, the sensitivity is about 2% lower at these points. A
typical energy correction algorithm measures the energy spectrum as a function
of position in the image using a source or sources with known energies. A linear
or higher order correction is then made to the estimated energy.
Spatial corrections are needed because of biases in estimated interaction
positions. These biases result, as described above, from discrepancies in the
formulas used to estimate the position and the actual behaviour. As mentioned,
these usually result in lines being bent towards space between PMTs. Typically,
separate corrections are made for the axial (y) and transaxial (x) directions. These
corrections involve imaging a mask with a grid of holes or lines in combination
with a flood source. This results in a series of bright spots or lines in the image
that correspond to the known positions of the holes or lines in the mask. A
function, typically a polynomial, is fit to the set of true points as a function of
the set of measured points. This function can then be used to correct a measured
position.
600
500
Counts
400
300
200
100
0
110
Average
2% lower
2% higher
120
130
140
150
Energy (keV)
160
170
Fig.11.18. Sample energy spectrum for 140keV photons for the cases of average, 2% lower
than average and 2% higher than average light collection efficiency. The variation in light
collection efficiency results in a shift of the energy spectrum, which results in non-uniform
sensitivity.
334
The final type of correction applied is a uniformity or sensitivity correction.

The goal of this correction is to make images of a flood source as uniform
as possible (seeFig.11.19). There are two types of uniformity corrections:
intrinsic, which corrects only for non-uniform sensitivity of the detector system
(i.e. excluding the collimator), and extrinsic, which corrects for both detector and
collimator non-uniformities. Uniformity corrections are made using a high-count
flood image. Uniformity correction is implemented by, in essence, multiplying
each pixel in acquired images by a factor equal to the average counts in the active
portion of the flood image divided by the counts in the corresponding pixel in the
flood image.
The number of counts in the flood image is critical in determining the
ultimate uniformity of the image. This is especially important in SPECT where
local non-uniformities can result in ring artefacts. To achieve this, the counts in
the flood image should be such that the relative standard deviation (coefficient
of variation) of the pixel counts resulting from Poisson counting statistics is less
than the desired uniformity. For example, if uniformity correction to better than
1% is desired, the average number of counts per pixel in the uniformity flood
should be greater than 1/0.012=10 kcounts per pixel. The total number of counts
in the flood, thus, depends on the number of active pixels in the image. Since
non-uniformities are generally relatively low frequency, this restriction can be
relaxed to some degree by the use of low pass filtering applied to the flood image.
Intrinsic flood images are usually acquired using a point (or syringe) source
containing a small quantity of the isotope of interest. If the source is placed at
a distance of more than five times the largest linear dimension of the camera
FIG.11.19. Intrinsic flood images for a gamma camera having a poor (left) and good
(right) set of corrections applied. It should be noted that the images are windowed so that
the brightness represents a relatively small range of count values in order to amplify the
differences. The peppery texture is due to quantum noise and is to be expected. The quantum
noise is exaggerated because of the windowing used, and can be reduced by acquiring very
high-count flood images.
335
CHAPTER 11
FOV from the camera face, then the irradiation of the camera can be considered
uniform. Since the uniformity of the camera will, in general, vary depending
on the energy of the isotope and energy window used, this correction should
ideally be made for each isotope and energy window used. The count rate for the
acquisition should be within acceptable limits to avoid high count rate effects.
Extrinsic flood images are made using a flood or sheet source. Fillable
flood sources have the advantage that they can be used for any isotope. However,
great care must be made in filling the phantom to remove bubbles, mix the
activity and maintain a constant source thickness. In addition, images must be
obtained for each collimator used with a given isotope. As a result, 57Co sheet
sources are often used to obtain extrinsic flood images. These have the advantage
of convenience but are, strictly speaking, valid for only a single isotope.
One way to take advantage of both approaches is to perform uniformity
corrections using a combination of intrinsic flood images for each isotope
used and an extrinsic flood image obtained for the collimator in question. This
approach assumes that collimator uniformity is independent of energy and can,
thus, be measured with, for example, a 57Co sheet source. Not all equipment
vendors support this approach. Some vendors assume that the energy and linearity
corrections produce uniformity that is energy independent, and they, thus,
recommend only the use of an extrinsic flood image for uniformity correction.
Another approach is to first confirm the uniformity of all collimators via a sheet
source flood image with intrinsic correction for 57Co. Then, an extrinsic flood
image for each isotope used is acquired and used in uniformity correction for that
isotope, assuming that the collimator is sufficiently uniform. The best approach
depends on the characteristics of the individual camera.
11.2.2.7. Image framing
The final step in generating gamma camera images is image framing.
Image framing refers to building spatial histograms of the counts as a function of
position and possibly other variables. This involves several steps, and is typically
done either by microprocessors in the camera or in an acquisition computer.
In this step, position is mapped to the elements in a 2D matrix of pixels. The
relationship between pixel index and physical position is linearly related to the
ratio of the maximum dimension of the FOV of the camera to the number of
pixels, the zoom factor and an image offset. The zoom factor allows enlarging
the image so that an object of interest fills the image. This can be useful, for
example, when imaging small objects. It results in a pixel size in the image that
is a factor of 1/zoom factor as large as in the unzoomed (zoom factor equals
one) image. It should be noted that even though the pixel size is decreased, the
resolution of the image will not necessarily be improved as long as the original
336
pixel size is smaller than the intrinsic resolution. For example, if the native pixel
size is 3.2 and the intrinsic resolution is 4mm, a zoom factor of two will result in
a pixel size of 1.6mm, but the intrinsic resolution will still be 4mm. An image
offset can be used to shift the image, so that an object of interest is in the centre
of the acquired image.
In addition to adding counts to the appropriate pixel spatially, the framing
algorithm performs a number of other important functions. The first is to reject
photons that lie outside of the energy window of interest. This is done to reject
scattered photons. Gamma cameras typically offer the ability to simultaneously
frame images corresponding to more than one energy window. This can be useful
for isotopes having multiple photopeaks, for acquiring data in energy windows
used by scatter compensation algorithms or for acquiring simultaneous images
of two or more radionuclides. Framing software typically enables the summation
of photons from multiple, discontiguous energy windows into one image as well
as simultaneously framing multiple images from different energy windows into
different images. There is often a limited number of energy windows that can
be framed into a single image, and a limit on the number of images that can be
framed at one time. These limits may depend on the image size, especially if the
framing is done by a microprocessor in the camera that has limited memory.
A second important function provided by the framing system is the ability to
obtain a sequence of dynamic images. This means that photons are recorded into
a set of images depending on the time after the start of acquisition. For example,
images could be acquired in a set of images with a frame duration of 10 s. Thus,
for the first 10 s, photons are recorded into the first image; for the second 10 s,
they are recorded into a second image; and so on. Thus, just as multiple images
are obtained in the case of a multi-energy window acquisition, multiple images
are obtained corresponding to a sequential set of time intervals. This is illustrated
in Fig.11.20, where seven dynamic frames are acquired for a time interval T.
Dynamic framing is used for monitoring processes such as kidney function,
gastric emptying, etc. The time frames are often not equal in duration as there
may be more rapid uptake at early times and a later washout phase in which the
change in activity with time is slower. The number or acquisition rate of dynamic
frames is often limited due to constraints in framing memory and this limit can
depend on the image size.
Gated acquisition is similar to dynamic acquisition in that photons are
recorded into a set of images depending on the time they are detected. However,
in gated acquisition, the time is relative to a physiological trigger, such as an
electrocardiogram (ECG) signal that provides a signal at the beginning of each
cardiac cycle. This is appropriate for processes that are periodic. The photons
are counted into a set of frames, each of which corresponds to a subinterval of
the time between the two triggers. For example, the bottom two illustrations
337
CHAPTER 11
in Fig.11.20 should be considered. In both cases, the interval between gates is

divided into four subintervals (in cardiac imaging, 8 or 16 frames are typically
used, but 4 are illustrated in this example for simplicity). The photons arriving
in each of the subintervals are counted in the corresponding frame. Thus, for
cardiac imaging, the activity distribution during the first quarter of the cardiac
cycle is imaged in frame 1, the second quarter in frame 2, etc. This is useful for
assessing wall motion and thickening. Just as in dynamic and multiple window
acquisitions, there may be limits on the size and number of frames due to limits
in framing memory on the acquisition computer.
Since the gate is derived from a physiological signal, and physiological
signals are not necessarily exactly periodic, the framing algorithm must handle
the case of variable time intervals between gate signals. In cardiac imaging,
this corresponds to variations in the heart rate. Figure11.20 also illustrates two
ways of dealing with this. In the first method, the frames correspond to fixed
time intervals. What is typically done is to measure the average heart rate and
to divide this by the number of frames per cycle. The photons arriving during
each of these time intervals are then binned into the image corresponding to
each time interval. Difficulty arises when there are variations in the length of the
cardiac cycle. Figure11.20 illustrates the case when there are beats that are 5%
longer and 5% shorter than average. For fixed time interval framing, the actual
interval for the fourth frame will be lengthened or shortened when the length of
the cardiac cycle changes. This will result in motion blurring of the gated images.
The alternative is to change the length of the subintervals for each beat based on
the time between gates for that beat. This eliminates the problem with motion
blurring described above. However, this also requires buffering of the events for
the period between gates before they can be framed. One additional advantage of
this approach is that bad beats (ones longer or shorter than the average beat by
more than a certain fraction of the average beat length) can be discarded.
The final acquisition mode is list-mode acquisition. In list-mode acquisition,
the energies and positions of incoming photons are simply saved to a file in the
order in which they appear. Additional information is recorded in the form of
events in the list-mode stream. These events include things such as physiological
triggers, gantry or table motion, start and stop of acquisition, and timing marks
which are injected at regular intervals. The advantage of list-mode data is that
they contain all of the information obtained during the acquisition. As a result,
it can be retrospectively reframed using different pixel sizes, energy windows,
frame intervals, etc. However, the downside is that list-mode files are very large
(typically eight or more bytes of data are stored for each photon). List-mode
is often not made available for routine clinical use, but can be very useful for
research use.
338
Dynamic acquisition
Time from start of acquistion

Gated acquisition-fixed frame intervals

Gated acquisition-fixed frame fractions

FIG.11.20. Comparison of dynamic and gated acquisition modes. In all cases, time is along
the horizontal axis. In dynamic acquisition, photons are framed into different images based
on the time after acquisition. In this case, the interval between dynamic frames is T. For
gated acquisition, photons are framed into a set of images based on the time in relation to
the previous gate signal. The gate signal is derived from a physiological trigger such as the
R wave in an electrocardiogram signal. For the two dynamic gating examples, the time interval
between gate signals 15 and 78 are the same, but the interval between gates 56 and 67 are
5% longer and 5% shorter than the average interval. Two methods of dividing the time interval
between gates, fixed intervals and fixed fractions are illustrated. In the fixed interval method,
the photons are framed into frames with fixed widths. In this case, the extra time (5% of T)
between gates 56 is not counted, and the interval into which photons are counted in frame 4
in the time interval between gates 67 is shortened by 20% (5% of T).
339
CHAPTER 11
11.2.2.8. Camera housing

The camera housing contains the radiation detector and provides a mount
for the collimators. The housing performs a number of important functions. First,
it must provide radiation shielding so that photons can only enter the camera
and be detected via the collimator. This shielding is made of lead and, since the
crystal and PMT array is large, the housing must also be large and is rather heavy.
In addition, the PMTs are very sensitive to magnetic fields. The housing, thus,
includes mu-metal shields around the PMTs to screen magnetic fields. Without
these, there can be variations in the sensitivity and uniformity due to changes in
the camera position relative to ambient magnetic fields, including the magnetic
field of the Earth. The PMTs and detection electronics are sensitive to variations
in temperature and generate a non-negligible amount of heat. As a result, the
housing must include a temperature control system, typically in the form of fans
to circulate air and provide ventilation.
A final important function of the camera housing is to provide mounting
for the collimators. High energy parallel and pinhole cameras can weigh more
than 100 kg, so the mounting system must be sufficiently strong to securely
support this much weight. The back face of the collimator (excluding pinhole
collimators) must be in close proximity to the crystal in order to provide the
highest possible resolution. Since collimators are often changed several times per
day, the mounting system must provide for easy collimator removal and change.
Most modern cameras include automatic or semi-automatic collimator exchange
systems. Thus, the collimator retaining system often includes an automated
locking system. Finally, modern cameras include touch sensors on the face of the
collimators and often include proximity sensors. The touch sensors activate when
the collimator face touches the patient or bed and disable camera motion. This is
done for patient safety and in order to avoid injuring the patient or damaging the
camera. A proximity sensor is sometimes also included. This typically consists
of an array of light emitting diodes with an opposing array of photodiodes. They
are mounted so that the light beam from the light emitting diodes is parallel to
the face of the camera. Proximity to the patient can be detected when one of the
light beams is interrupted by part of the patient. The proximity sensor can be used
for automated sensing of the camera orbit that positions the cameras close to the
patient at each camera position. Electrical connections between the housing and
the collimator provide communication with the touch and proximity sensors as
well as providing information about which collimator type is mounted.
340
FIG.11.21. (a)A cross-section of a dual head gamma camera capable of acquiring two views
simultaneously. It should be noted in this example that the heads are oriented 180 apart,
although a 90 degree configuration is also possible. SPECT data acquisition requires rotation
of the gamma camera heads about the long axis of the patient as indicated. (b)A transverse
slice with the position of four different camera orientations superimposed to illustrate the
acquisition of multiple angular views.
11.2.3. SPECT
11.2.3.1. Gamma camera SPECT systems
In addition to the software requirements for image reconstruction, SPECT
is associated with hardware requirements that are beyond those needed for
planar imaging. Although various SPECT configurations have been developed,
the most common implementation involves the use of a conventional gamma
camera in conjunction with a gantry that allows rotation of the entire detector
head about the patient [11.2]. The gantry rotation is about the long axis of the
patient (seeFig.11.21) and is typically performed in discrete steps (step and
shoot), although continuous motion may also be supported. During rotation of
the gantry, the patient bed typically does not move, so SPECT data acquisition
is more similar to conventional CT than to spiral CT, in this respect. In this way,
planar views of in vivo radioactivity distribution can be acquired at different
angular orientations and these data can be used to form the projections that are
required for image reconstruction by computed tomography.
341
CHAPTER 11
In principle, rotation of the gamma camera about 180 allows for the
acquisition of sufficient projections for tomographic reconstruction. However,
in practice, opposing views acquired 180 apart differ due to various factors
(photon attenuation, depth dependent collimator response) and SPECT data are
commonly acquired over 360. The theory of computed tomography determines
the number of angular samples that are required, but for many SPECT studies,
around 128 views may be acceptable. The time needed to acquire these multiple
projections with adequate statistical quality is a practical problem for clinical
SPECT where patient motion places a limit on the time available for data
acquisition. In an effort to address this issue, a common approach in modern
SPECT designs is to increase the number of detector heads, so that multiple
views can be acquired simultaneously. Dual detector head systems currently
predominate, although triple detector head gamma cameras also exist. Increasing
the number of detector heads increases the effective sensitivity of the system for
SPECT, at the expense of increasing cost. Dual head gamma cameras are often
considered the preferred configuration for systems intended not just for SPECT
but also for general purpose applications, including whole body studies where
simultaneous acquisition of anterior and posterior planar images is required.
In addition to the rotational motion required for SPECT, flexibility is
also required in the relative positioning of the detector heads. For general
purpose SPECT with a dual head system, the two heads are typically oriented
in an opposing fashion (sometimes referred to as H-mode) and 360 sampling
is achieved by rotation of the gantry through 180. In contrast, cardiac SPECT
is often performed with the detectors oriented at 90 to each other (sometimes
referred to as L-mode). In this mode, the gantry rotates through 90 and the two
detectors acquire projections about 180 from the right anterior oblique position to
the left posterior oblique position. Despite acquiring only 180 of data, this mode
has advantages for cardiac applications as it canminimize the distance between
the heart and the detectors, thus reducing attenuation and depth dependent losses
in spatial resolution. Other approaches tominimizing the distance between the
detectors and the patient during SPECT data acquisition involve further control of
the rotational motion of the detector heads. For detectors rotating about a circular
orbit, this involves adjusting the radius of rotation for individual studies, so as
tominimize the source to collimator distance. Other options include detectors
that rotate, not in a circular orbit, but in an elliptical orbit, or, alternatively, a
variable rotational motion that contours to the outline of the body.
The flexibility of the motions that are available in modern SPECT systems
makes it particularly important to ensure that the detectors are correctly aligned.
This means that the specified angle of rotation is accurately achieved at all angles.
The detector heads also need to be perfectly oriented parallel to the z axis of the
system, such that each angular view is imaging the same volume. Furthermore,
342
it is important that the centre of each angular projection is consistent with the
centre of mechanical rotation. Errors due to these factors can potentially lead
to a loss of spatial resolution and the introduction of image distortion or ring
artefacts. In order to identify and correct these issues, an experimental centre of
rotation procedure is employed. A small point source is placed in the FOV at an
off-centre location. SPECT data acquisition is performed and deviations from the
expected sinusoidal pattern are measured in the resulting sinograms.
FIG.11.22. (a)A series of planar views acquired at different angular orientations. A sample of
four views has been extracted from a total of 128 views acquired about 360. It should be noted
that the z axis represents the axial position and is the axis of gantry rotation. (b)A sinogram
corresponding to a particular axial location. The red lines in (b)indicate 1D projections that
have been extracted from the corresponding planar views shown in (a).
Although SPECT can be performed with a variety of collimator geometries,

such as cone-beam or pinhole, much of the present discussion has assumed
parallel-hole collimation. In this case, each planar view consists of multiple
1D projections, each measured at different axial positions (Fig.11.22). Each
projection is defined by the holes of the collimator and approximates a series of
parallel line integrals of the activity distribution in the FOV. In practice, these line
integrals are substantially corrupted by the effects of photon attenuation, scatter
and depth dependent collimator response. Each of these factors requires software
correction and these corrections are described in the following paragraphs.
343
CHAPTER 11
11.2.3.2. Attenuation correction

Standard tomographic reconstruction algorithms, such as those based on
filtered back projection, assume that measured projections are line integrals
through the object. However, in SPECT, the interaction of photons via
photoelectric absorption and Compton scatter within the patient results in
attenuated projections. The attenuated projections P(t) can be described for the
2-D case by the equation:
P (t ) =
a(ln + tm )e
(l 'n +tm ) dl '

0
dl
(11.8)
where the geometry is illustrated in Fig 11.23. In this equation, the unit vectors
n and m are as described in the legend of Fig. 11.23, t is the transaxial distance
in the projection from the projected position of the origin and l is the distance
along the projection line from the face of the detector. a(x) is the activity
distribution and gives the activity at point x. It should be noted that the integral
in the exponent represents the integral through the attenuation distribution (x)
from the point x = ln + tm. Thus, the exponential represents the attenuation of
photons emitted at x as they travel back towards the detector.
FIG. 11.23. Projection geometry used to describe the attenuated projection in Eq. (11.8). In
this figure, the projection is at an angle . A parallel-hole collimator is assumed, and the unit
vector n is perpendicular to the collimator and parallel to the projection rays. The unit vector
m is parallel to the collimator face and perpendicular to n. The variable t is the distance
along the detector from the projected position of the origin.
344
As can be seen from the above equation, unlike PET, the attenuation
is not constant for a projection ray, but instead varies along the ray. Using
reconstruction methods that do not model this effect produces both artefacts and
a loss of quantitative accuracy in the resulting images. The artefacts can include
streak artefacts, resulting from highly attenuating objects such as bones, catheters
or medical devices; shadows, due to higher attenuation between an object in
some views than in others (e.g. breast or diaphragm artefacts in cardiac SPECT);
and a generally reduced image intensity in the centre of the image.
The first requirement to compensate for attenuation is knowledge of the
attenuation distribution in the patient. This is done by either assuming uniform
attenuation inside the object and extracting information about the body outline
from the emission data or using a direct transmission measurement. Assuming
a uniform attenuation distribution in the patient is only valid in regions such as
the head. Even in the head, bone and regions containing air, such as the sinuses,
result in imperfect estimates of the attenuation distribution and lead to imperfect
attenuation compensation. Myocardial perfusion imaging is an important
application for SPECT and, since attenuation can produce artefacts that obscure
actual perfusion defects, a number of commercial devices have been developed
to allow measurement of the attenuation distribution in the body. All of these
devices use transmission CT techniques to reconstruct the attenuation distribution
inside the body. The devices that have been developed can be divided into two
general classes: devices using radionuclide sources and devices based on X ray
tube sources. In both cases, a source of X rays or radiation is aimed at the body
and a detector on the opposite side of the body measures the transmitted intensity.
The intensity I(t) passing through the body for a source with incident intensity
I0, projection position t and projection view is given by:
I (t ) = I 0 (t )e
(ln +tm ) dl
0
(11.9)
where the symbols and geometry are as in Fig. 11.23.

Acquiring sets of these transmission data for various angles allows
reconstruction of the attenuation distribution. Tomographic reconstruction
methods can be applied directly by noting that the negative of the log of
the fraction of transmitted photons is a line integral through the attenuation
distribution.
A number of transmission devices based on radionuclide sources have been
developed and marketed (Fig.11.24). All of these devices use the gamma camera
to detect the transmission photons. The simplest of these designs is a sheet
source of radioactivity. To avoid contaminating the projection data, either the
345
CHAPtER 11
transmission scan is acquired separately from the emission data or simultaneously

using a radionuclide with a lower photopeak energy. Typically, 153Gd is used as
it has an energy lower than that of 99mTc and the transmission photons, thus, do
not interfere with collection of emission data. To reduce patient dose and scatter
in the transmission data, the source should be collimated. The disadvantages of
sheet source designs are that they are expensive and high activities make them
dangerous to handle. using lower activity sources results in contamination of the
transmission data.
FIG. 11.24. Illustration of a number of proposed transmission scanning devices. It should be

noted that in all cases the drawing represents a sagittal view of the system. However, for (d),
the multiple line source system, the line sources are normally parallel to the long axis of the
patient and, thus, perpendicular to the direction shown. Similarly, for (c), the fan-beam system,
the fan beam is in the transaxial direction, the opposite of that pictured above. Furthermore,
for the sheet source system, the source is continuous and the fan of rays is only shown to
illustrate the limited range of directions of transmission source photons that pass through the
source collimator.
as a result of these difficulties, line source transmission systems have been

developed. The line sources are either scanned opposite the patient and parallel
to the collimator or positioned at the focal spot of a fan-beam collimator. in the
latter case, the projection geometry is different from the parallel-beam geometry
346
described above, but the same basic principles apply. The fan-beam geometry
has the advantage of having much higher sensitivity and, thus, of requiring lower
activity sources. However, the fan-beam geometry results in magnification and,
thus, the FOV is smaller than the size of the detector. It should be noted that in the
fan-beam geometry the fan lies in the transaxial plane and, thus, the truncation is
in the transaxial direction. For the scanning line source systems, an electronic
window is used so that transmission data are acquired only in the region directly
under the line source. To overcome the low sensitivity of parallel-beam geometry,
high activities (1.8518.5 GBq) are used. This reduces the contamination of the
transmission data by emission activity. The scanning line source adds additional
mechanical complexity to the system and, as a result, several other designs have
been proposed.
In multiple line source systems, a small number of line sources are used.
The line sources are placed close enough together so that the object is covered
due to the finite acceptance angle of the camera collimator, in effect acting like a
non-uniform sheet source. One advantage of this geometry is that weaker sources
can be used for the sources near the edge of the camera, since these correspond
to thinner regions of the body. This can help reduce the effects of high count
rates. The final geometry proposed is the half-cone geometry shown above. In
this design, a high energy 133Ba point source is used. Many of the high energy
photons penetrate through the collimator, creating a half-cone-beam geometry.
This allows a parallel-hole collimator to be used both for transmission and
emission imaging. The use of a point source simplifies shielding of the source.
In general, radionuclide transmission sources have a number of
disadvantages. These include the fact that the source decays and must be replaced.
There are also limits on transmission count rates imposed by the gamma camera,
resulting in relatively noisy transmission images. In addition, if the count rate
due to the emission activity within the patient is high, the transmission images
can be degraded, resulting in inaccurate attenuation maps. Finally, with the
sheet source, scanning line source and multiple line source geometries, the
resolution of the transmission scan is limited by the combination of source and
camera collimators. In general, these provide lower resolution transmission
scans. However, one advantage of radionuclide based transmission systems is
the potential to perform simultaneous imaging, thus eliminating the need for
an additional transmission scan. Another advantage, especially when acquired
simultaneously, is that registration of the emission and transmission images is
guaranteed. Finally, the use of radionuclide sources with a small number of high
energy photopeaks makes converting the transmission images into an attenuation
map at the energy of the emission source easier than for X ray CT based systems,
which use X ray tubes having continuous X ray energy spectra.
347
CHAPTER 11
The second major kind of transmission scanning apparatus uses X ray

sources. There are two major kinds: slow rotation and hybrid SPECT/CT
systems. In the slow rotation devices, a low power X ray tube is attached to the
gantry opposite a conventional X ray detector. The gantry is rotated around the
patient, resulting in acquisition of X ray transmission data. In one commercially
available slow-rotation device, the X ray detector has 14 rows of detectors
with an effective axial length of 0.51.0cm at the centre of rotation and can,
thus, acquire transmission data for 14 slices per revolution. The patient is then
translated on the bed to acquire data from additional slices. The resolution of
these scanners can be as good as 1mm transaxially but is 0.51cm axially.
Typical revolution times for this type of scanner are 2030 s/rotation. Another
commercial implementation uses a flat panel detector and cone-beam geometry.
This provides very high resolution, though there is the potential for cone-beam
artefacts in the reconstructed images. In hybrid SPECT/CT systems, a SPECT
gantry and an independent diagnostic CT system are integrated. The CT images
are acquired very rapidly using conventional CT technology either just before
or just after acquisition of the SPECT data. The CT images acquired using these
systems, as well as the acquisition protocols and options, are similar to those on
diagnostic CT scanners.
X ray tube based methods for obtaining transmission images have a
number of advantages and disadvantages. The major advantages are acquisition
speed, the quality of the attenuation maps (high resolution and low noise) and
the convenience of not needing to replace radionuclide sources. Among the
disadvantages are that the image is not acquired simultaneously and is often
acquired with the bed in a different position than used for the SPECT scan. There
is, thus, the potential for mis-registration of the SPECT images and attenuation
maps, resulting in degraded attenuation compensation. A second disadvantage is
that the effects of motion, especially respiratory motion, during the attenuation
scan are different to those during the emission scan. For example, in slow-rotation
devices, streak artefacts in the diaphragm region caused by respiratory motion
during transmission data acquisition are common. For hybrid devices, the CT
scans are acquired very rapidly, so may freeze the patient at one particular phase
of the respiratory cycle. Despite these potential disadvantages, the advantages of
X ray tube based transmission scanning (including hybrid SPECT/CT systems)
have meant that they have largely replaced devices based on radionuclide sources.
Using the above systems for attenuation compensation requires
transforming the transmission CT image into an emission attenuation map. For
radionuclide sources, this basically means translating the attenuation coefficients
at the energy of the transmission source to that of the emission source. For
emission and transmission source energies above 100keV, this can be done by
scaling the transmission CT image by the ratio of the attenuation coefficient in
348
water at the emission energy divided by that at the transmission source energy.
However, more accurate results can be obtained using the bilinear scaling method
described below for use with X ray CT images.
For hybrid SPECT/CT systems where a conventional X ray CT image
is transformed into an attenuation map, there are a number of additional
considerations. First, transmission data are obtained using a polychromatic
source. There can, thus, be substantial beam hardening. Fortunately, beam
hardening and other corrections routinely applied in X ray CT scanners to produce
images in Hounsfield units (HU) eliminate many of these concerns. In this case,
the CT image in Hounsfield units can be transformed to the attenuation map via
piecewise linear scaling, where, in effect, pixels with values less than 0 HU are
treated as water with densities ranging from 0 to 1, pixel values between 0 and
1000 HU are treated as a mixture of bone and water, and pixel values greater
than 1000 HU are treated as dense bone. Thus, for a pixel having a valueh in
Hounsfield units, the attenuation map value is given by:
1000 + h
water
for h 0
1000
h
(h) = water +
water ) for 0 < h < hbone (11.10)
(
hbone bone
h
bone
for h > hbone
hbone
where water and bone are the attenuation coefficients of water and bone,
respectively, for the energy of the photopeak of the imaging radionuclide.
Once the attenuation map is obtained, attenuation correction can be
implemented using analytical, approximate or statistical image reconstruction
algorithms. Generally, analytical methods are not used due to their poor noise
properties. Approximate methods include the Chang algorithm. This method is
often used in regions of the body where the attenuation coefficient is assumed
to be uniform and the actual attenuation distribution is not measured, but instead
is approximated from the boundary of the object, which is usually assumed to
be an ellipse. In the Chang method, an image reconstructed using filtered back
projection (i.e. without attenuation compensation) is approximately compensated
for attenuation. This approximate compensation is obtained for each voxel by
dividing the uncorrected image signal by the average of the attenuation factors
that correspond to each projection view. For a uniform attenuator with an assumed
elliptical boundary, these attenuation factors can be calculated analytically.
However, the Chang method is approximate and has poor noise properties.
349
CHAPTER 11
Thus, for the best attenuation compensation, statistical iterative reconstruction

methods should be used. Statistical iterative reconstruction methods are discussed
in more detail in Chapter13. These methods can be used to compensate for
attenuation by incorporating a model of the attenuation process in the imaging
matrix. Accurate attenuation compensation can be obtained with fast statistical
iterative reconstruction methods, such as the ordered-subsets expectationmaximization (OSEM) algorithm, if an accurate attenuation map is available.
11.2.3.3. Scatter correction
In gamma camera imaging, a significant fraction of the detected photons
are scattered in the body. This is due to the finite energy resolution of the gamma
camera, which results in imperfect energy based scatter rejection. The scatter
to primary ratio (SPR) depends on the radionuclide, energy window, energy
resolution, source depth and the size of the object. For example, for 99mTc, the
SPR is in the range of 0.2 for brain imaging and 0.6 for cardiac imaging. On
the other hand, for 201Tl cardiac imaging, the SPR can be greater than 1. Scatter
results in loss of contrast, especially for cold objects in a warm background, and
loss of quantitative accuracy.
Scatter correction requires estimating the scatter component of the
projection data combined with a compensation method. Most frequently, the
scatter component is estimated using data acquired in auxiliary energy windows.
Perhaps the most common and flexible such method is the triple energy
window (TEW) method. This method uses two scatter energy windows, one
above and one below the photopeak window, as illustrated in Fig.11.25. The
scatter is estimated from the counts in the scatter windows using a trapezoidal
approximation where counts in the scatter windows divided by their window
widths are treated as the sides, and the scatter in the photopeak window is the
area of the trapezoid. The estimated scatter counts in the photopeak window
estimated using TEW sTEW are given by:
c
c upper w peak
s TEW = lower +
(11.11)
w lower w upper 2
where
are the counts in the lower and upper scatter windows,
clower and cupper
respectively;
and wpeak, wlower and wupper are the widths of the photopeak, lower scatter and
upper scatter windows, respectively.
350
It should be noted that in Fig.11.25 the scatter windows are adjacent to the
photopeak energy window, but this is not necessary, nor, in all cases, desirable. In
fact, it is desirable to position the windows as close as possible to the photopeak
window, while having only a small fraction of the photopeak photons detected in
the scatter windows. The width of the scatter windows is a compromise between
obtaining as accurate a scatter estimate as possible, which favours a narrow
energy window, but having an estimate that is low in noise, which favours a wide
energy window.
W
FIG.11.25. Illustration of the use of a trapezoidal approximation to estimate the scatter in the
photopeak energy window in the triple energy window method scatter compensation for 99mTc.
It should be noted that in this example the windows are not necessarily optimally placed. In
particular, the scatter windows are positioned such that there is a non-zero contribution from
unscattered photons in the scatter energy windows. This is especially evident for the upper
energy window. For the case of 99mTc, the counts in the upper window are often assumed to be
zero.
Another method to estimate the scatter component in the projection data

is via the use of scatter modelling techniques. These techniques use an estimate
of the activity distribution and a mathematical algorithm to compute the scatter
that would have been detected from the activity distribution. The mathematical
techniques used range from accurate approximations to full Monte Carlo
simulations.
As mentioned above, scatter correction is accomplished by combining
scatter estimation and compensation methods. Methods of compensating for
scatter include subtracting the scatter estimate from the projection data and, for
351
CHAPTER 11
SPECT, including the projection data in the iterative reconstruction process.

Owing to noise and errors in the scatter estimate, subtraction of a scatter estimate
from measured projection data can lead to some negative pixel values. This
can be a problem when the data are subsequently used with statistical iterative
reconstruction algorithms. For scatter estimates obtained from energy windows,
this effect can be reduced by low pass filtering the scatter estimate. Nevertheless,
truncation of negatives is often used, though this can increase bias in the scatter
compensation.
For SPECT, a better way to accomplish scatter compensation is to add the
scatter estimate to the computed projection during the iterative reconstruction
process. For example, algorithms such as OSEM and maximum-likelihood
expectation-maximization (MLEM) involve back projecting the ratio of the
measured and computed projections. In this case, the scatter estimate can simply
be added to the computed projections. Another approach is to include scatter
modelling in the projection matrix. Either of these approaches is superior to
pre-subtraction of the scatter estimate.
11.2.3.4. Collimator response compensation
Images obtained with a gamma camera are degraded by the spatially
varying collimatordetector response (CDR). For parallel-hole collimators, the
CDR depends approximately linearly on the distance from the collimator face.
The CDR has geometric, septal penetration and septal scatter components.
These correspond, respectively, to photons passing through the collimator holes,
photons passing through the septa without interacting, and photons scattering in
the septa and resulting in a detected photon. The latter two effects tend to reduce
image contrast, can produce star artefacts and introduce distance dependent
sensitivity variations. Septal penetration and scatter can be reduced with the
use of a properly designed collimator. However, for isotopes emitting photons
with energies greater than ~300keV, including isotopes having low abundance
high energy photons not used for imaging, some level of these effects is almost
unavoidable.
Since SPECT images contain information about the distance from the source
to the collimator, it is possible to provide improved compensation for the CDR
as compared to planar imaging. This can be accomplished using both analytical
and iterative methods. However, analytical methods involve approximations and
usually have suboptimal noise properties and are, generally, not commercially
available. As a result, methods based on statistical iterative reconstruction have
been developed and are commercially available. These methods model the CDR
in the projection and back projection process. In SPECT, the matrix elements
are not explicitly calculated and stored, but are implicitly calculated during
352
the reconstruction using a projector and back projector algorithm. Thus, CDR
compensation is accomplished by modelling the CDR in the projector and back
projector. This is often implemented by rotating the image estimate, so that it is
parallel to the collimator face at each projection view. In this orientation, the CDR
is constant in planes parallel to the collimator face and can, thus, be modelled by
convolution of the CDR for the corresponding distance. In order to do this, the
distance from the plane to the face of the detector is needed. This is somewhat
complicated by the use of non-circular orbits and, in this case, manufacturers
need to store the orbit information (distance from the collimator face to the centre
of rotation for each projection view) with the projection image. In addition, a way
of estimating the CDR is needed. Analytical formulas exist for calculating the
geometric component of the CDR. Alternatively, a Gaussian function fit to a set
of measured point response functions measured in air can be used. Compensation
for the full CDR, including septal penetration and scatter, requires a CDR that
includes these effects. Analytical formulas do not exist, so either numerically
calculated (e.g. using Monte Carlo simulations of the collimatordetector system)
or measured CDRs are used. Various optimization and speed-up techniques have
been implemented to reduce the time required for CDR modelling.
It should be noted that CDR compensation does not fully recover the loss
of resolution of the collimator: the resolution remains limited and spatially
varying and partial volume effects are still significant for small objects. In
addition, CDR compensation results in correlated noise that can give a blobby
texture to the images (though the images do indeed seem qualitatively less
noisy), and results in ringing artefacts at sharp edges. Despite these limitations,
CDR compensation has generally been shown to improve image quality for both
detection and quantitative tasks and has been used as a way to allow reduced
acquisition time.
11.3. PET SYSTEMS
11.3.1. Principle of annihilation coincidence detection
Radioactive decay via positron emission is at the heart of the PET image
formation process [11.3]. Positrons are emitted from the nucleus during the
radioactive decay of certain unstable, proton-rich isotopes. These isotopes
achieve stability by a decay process that converts a proton to a neutron and is
associated with the creation of a positron. A positron is the antimatter conjugate
of an electron and has the same mass as an electron but positive charge. As
with decay, positrons are emitted from the nucleus with different energies.
These energies have a continuous spectrum and a specific maximum value
353
CHAPTER 11
that is characteristic of the parent isotope. Once emitted from the nucleus, the
positron propagates through the surrounding material and undergoes scattering
interactions, changing its direction and losing kinetic energy (Fig.11.26). Within
a short distance, the positron comes to rest and combines with an electron
from the surrounding matter. This distance is dependent on the energy of the
positron, which is itself a function of the parent isotope and is typically on the
order of a millimetre. The combination of a positron and an electron results in
the annihilation of both particles and the creation of two photons, each with an
energy of 511keV, equivalent to the rest masses of the two original particles.
Conservation of momentum, which is close to zero immediately before
annihilation, ensures both photons are emitted almost exactly 180 apart. These
characteristic photon emissions (known as annihilation radiation) always
511keV, always emitted simultaneously and almost exactly 180 apart form
the basis of PET and result in distinct advantages over single photon imaging in
terms of defining the LOR.
Positronelectron
annihilation
FIG.11.26. Positrons emitted from a radioactive nucleus propagate through the surrounding
material before eventually coming to rest a short distance from their site of emission. At this
point, the positron annihilates with an electron, creating two 511keV photons that are emitted
approximately 180 apart. The perpendicular distance from the line defined by the two photons
to the site of positron emission places a limit on the spatial resolution that can be achieved
with PET systems.
The advantages of PET over SPECT, in terms of improved spatial

resolution, statistical quality and quantitative accuracy, can be attributed to the
fact that PET does not require a collimator and, therefore, eliminates the weakest
link in the SPECT image formation process. Instead of physical collimation,
354
PET systems employ a form of detection that can be thought of as electronic

collimation. If a positron source is surrounded by suitable detectors, both back
to back photons from an individual positron decay can potentially be detected
(Fig.11.27). As both photons are emitted simultaneously, they will be detected
at approximately the same time, allowing temporal acceptance criteria to be used
to associate pairs of corresponding detection events. This mode of detection is
referred to as coincidence detection and allows corresponding photon pairs to
be distinguished from other unrelated, potentially numerous, photon detection
events. As both photons that arise from positron decay are emitted almost exactly
180 apart, coincidence detection can be used to help localize the source of the
photon emissions. In general, a line drawn between corresponding detectors can
be assumed to intersect the point of photon emission, although information is
usually not available about exactly where along that line the emission occurred.
However, if a system of detectors is arranged at different positions around
the source, multiple coincidence events can be recorded at different angular
orientations. Over the course of an extended scanning period, a large number
of coincidence events will be recorded and angular projections of the activity
distribution can be estimated. These projections may then be used to reconstruct
3D images using the methods of computed tomography.
FIG.11.27. The back to back photons that result from positronelectron annihilation can
potentially be measured by detectors placed around the source. Coincidence detection involves
the association of detection events occurring at two opposing detectors (A and B) based on
the arrival times of the two photons. A line of response joining the two detectors is assumed to
intersect the unknown location of the annihilation event. Coincidence detection obviates the
need for a collimator and is sometimes referred to as electronic collimation.
355
CHAPTER 11
11.3.2. Design considerations for PET systems

11.3.2.1. Spatial resolution
High spatial resolution is clearly an important design objective for PET
imaging systems. As such, the trend in modern scanner systems has been to
decrease the width of individual detectors and to increase the total number of
detector elements surrounding the patient. The increased concentration of
detector elements decreases the sampling interval and generally improves spatial
resolution. Although modern designs involve detectors that are only a few
millimetres wide, the need for high sensitivity means that they are often a few
centimetres long. Problems can occur when photons are incident on one detector
but penetrate through to an adjacent detector (Fig.11.28(a)). The location within
the detector (depth of interaction) is typically not measured and the detection
event is assigned to a location at the face of the detector. This problem frequently
occurs when detectors are arranged in a ring (or similar) configuration and gives
rise to a loss of resolution at more peripheral locations. This resolution loss
generally occurs in the radial direction as opposed to the tangential direction due
to the angle of incidence of the photons on the detectors.
Radial
resolution
Tangential
resolution
FIG.11.28. (a)Photon penetration between adjacent detectors in a ring based system leads
to mis-positioning of events. This primarily affects the radial component of spatial resolution
which degrades with distance from the centre of the field of view. (b) Residual momentum
of the positron and electron immediately before annihilation causes the two 511keV photons
to deviate slightly from the expected 180 angle. As a result, a line joining detection events
does not intersect the exact point of annihilation. The extent of this non-collinearity is greatly
exaggerated in the figure, but it does contribute to a loss of spatial resolution, especially for
large diameter PET systems.
356
Another factor that influences spatial resolution is the distance between

opposing detectors. This distance is relevant because of a small uncertainty in
the relative angle of the 511keV annihilation radiation (Fig.11.28(b)). Although
the basic assumption of coincidence detection is that annihilation radiation is
emitted 180 apart, this is not strictly true. Positrons frequently annihilate before
they have lost all momentum, and this residual momentum translates to a small
deviation of about 0.25 from the expected back to back emissions. This effect is
referred to as non-collinearity and tends to degrade spatial resolution as detector
separation increases. For PET systems with opposing detectors separated by
only a few centimetres, such as those optimized for specific organs such as the
brain or breast, this is not a major issue. However, for whole body systems, in
which opposing detectors are typically separated by about 80cm, the effect of
non-collinear photons contributes a blurring with an FWHM of approximately
2mm.
The distance travelled by a positron between its point of emission and
annihilation is an additional factor that degrades the spatial resolution that can
be achieved by PET systems. As previously discussed, this distance, or positron
range, is dependent on the energy of the positron and also the type of material
through which the positron is passing; a greater range is expected in tissue such
as lung compared to soft tissue. It should be noted that software corrections have
been implemented that model the effect of positron range and can potentially
reduce the loss of resolution in reconstructed images.
11.3.2.2. Sensitivity
The best possible spatial resolution that can be obtained by a PET system is
not always achieved in clinical practice due to statistical noise in the measured
data (Fig.11.29). In order to suppress this noise, clinical protocols generally
employ low-pass filters or other similar image reconstruction methods, but the
consequence is invariably a loss of spatial resolution. Improving the statistical
Fig.11.29. Images of the same phantom, each showing different statistical quality. The images
shown in (a), (b), (c), (d), (e)and (f)were acquired for 1, 2, 3, 4, 5 and 20min, respectively.
Increasing the acquisition time increases the total number of true coincidence events and
reduces statistical variability in the image.
357
CHAPTER 11
quality of the measured coincidence data not only reduces image noise but also
allows the opportunity to reduce image smoothing and improve spatial resolution.
The need to optimize this trade-off between statistical noise and spatial resolution
influences both image reconstruction development and scanner design.
Noise in PET images is influenced by a number of factors, including the
sensitivity of the detector system, the amount of radioactive tracer administered
to the patient and the amount of time the patient can remain motionless for
an imaging procedure. Limitations on the latter two factors mean that high
sensitivity is an important objective for scanner design. Sensitivity is determined
by the geometry of the detector arrangement and the absorption efficiency of
the detectors themselves. Reducing the distance between opposing detectors
increases the solid angle of acceptance and increases sensitivity. However, the
requirement to accommodate all regions of the body imposes aminimum ring
diameter for whole body systems. For such systems, extending the axial field
of view (AFOV) of the detector system provides a mechanism for increasing
sensitivity. Cost constraints have prevented the construction of PET systems
that cover the entire length of the body. However, extended axial coverage can
be achieved using systems with much smaller AFOVs by scanning sections
of the body in a sequential fashion. For such systems, extending the AFOV of
the scanner not only increases sensitivity but also reduces the number of bed
translations required for whole body coverage. In addition to the geometry of the
detector system, sensitivity is also determined by the absorption efficiency of the
detectors. A high absorption efficiency for 511keV photons is desirable in order
to make best use of those photons that are incident upon the detectors. Absorption
efficiency or stopping power of the detector material is, therefore, an important
consideration for PET system design.
11.3.2.3. Quantitative accuracy
One of the strengths of PET is its capability to quantify physiological
processes in vivo. A prerequisite for this kind of quantitative analysis is that the
images accurately reflect the local activity concentration in the body. In order to
ensure this kind of quantitative accuracy, it is important tominimize effects that
corrupt the data and to correct residual corruption as necessary. Quantitative error
can arise from many sources but is primarily due to random coincidence events,
photon scatter within the body, photon attenuation within the body and detector
dead time. Figure11.30 illustrates some of these situations.
Figure11.30(b)illustrates a type of unwanted coincidence event that can
occur when photons from unrelated annihilation events are detected at
approximately the same time. Two photons detected within a short time interval
(coincidence timing window) will be associated with each other under the
358
assumption that they originated from the same positronelectron annihilation.

However, when there is a large amount of radioactivity within the FOV, it is very
possible that photons from unrelated annihilation events will be detected within
this time interval, leading to a spurious random coincidence. Figure11.30(c)
illustrates the case where a scattered photon is deflected from its original direction
but still reaches the detectors. In this case, a coincidence event can potentially be
recorded between the scattered and unscattered photons, leading to an inaccurate
coincidence event whose LOR does not pass through the true location of the
original annihilation. If not adequately corrected, such scattered coincidences
contribute a spurious background to the image that is dependent upon the size
and composition of the patients body and is non-uniform across the FOV. In
Fig.11.30(d), one of the two annihilation photons has been attenuated within the
body and only one photon is detected. No coincidence event is possible and the
FIG.11.30. (a) A true coincidence event can occur when both photons escape the body
without interacting. (b)A random coincidence event occurs when two photons from unrelated
annihilation events are detected at approximately the same time. (c)A scattered coincidence
event can occur when either photon is scattered within the body but is still detected. (d)No
coincidence event is recorded when one or both photons are attenuated, typically due to scatter
out of the field.
359
CHAPTER 11
scanner underestimates the signal that would be measured in the absence of

attenuation. The degree of underestimation depends on the size and distribution
of the patients body and results in quantitative error and gross image artefacts
that are non-uniform across the FOV. Attenuation can occur when one of the two
photons undergoes photoelectric absorption within the body. However, a more
likely occurrence is that one of the two photons is scattered within the body and
is deflected out of the FOV, leaving only one photon to be detected. Another
problem that can occur when large amounts of radioactivity are present is that
true coincidence events can potentially be lost due to the limited count rate
capability of the detection system. Each component of the system requires a finite
amount of time to process each event and, if another photon is detected during
this time, the second event will be lost. This dead time effect becomes significant
at high count rates and leads to an underestimation of the local activity
concentration in the images.
Software processing prior to (or during) image reconstruction can mitigate
the above effects, but the accuracy of these corrections may not be reliable if the
contamination overwhelms the signal from true coincidence events. PET systems
are, therefore, designed tominimize the contribution of the various degrading
factors described above. In terms of scanner design, very little can be done to
reduce attenuation as photons that are absorbed within the body do not reach
the detectors. However, scattered photons can potentially be rejected by the
detection system if their energy falls outside a predetermined acceptance range.
Annihilation radiation that is scattered within the body will emerge with energies
less than 511keV. The exact energy of the scattered photon will depend on the
number of Compton scattering interactions that have occurred and the angle
through which the photon was scattered. Energy discrimination can, therefore,
be used to reject photons that have energies less than 511keV and are assumed
to have been scattered. This approach relies on the detection system having high
energy resolution and, in practice, energy discrimination reduces but does not
eliminate scattered coincidence events. The limited energy resolution of current
PET systems means that, in order to avoid rejecting too many true (unscattered)
coincidence events, the energy acceptance window is usually set to accept quite a
broad range of energies around 511keV. High energy resolution is, nevertheless,
an important design objective, particularly for those scanner systems that detect a
large proportion of scattered photons.
Decreasing the coincidence timing window decreases the number of
random events as the shorter time interval reduces the likelihood of a coincidence
occurring by chance between two unrelated photons. There is, however, only
limited scope for reducing the duration of the coincidence time window as
it is restricted by the timing resolution of the detector system and the fact that
off-centre annihilations result in real time differences between detection events.
360
Optimization of the coincidence timing window for a particular scanner represents

a compromise between wanting to reduce the number of random coincidence
events without significantly reducing the number of true coincidences. Detector
systems that are able to measure photon detection times with low variability
(high timing resolution) are, therefore, desirable from the perspective of randoms
reduction. Timing resolution also contributes to detector dead time as a shorter
coincidence timing window reduces the likelihood of more than two photon
detection events occurring. Other contributions to dead time include the time
required by the detector to measure an individual photon event and the time spent
processing coincidence events.
11.3.2.4. Other considerations
Spatial resolution, sensitivity and quantitative accuracy are the main physics
issues influencing PET system design, but there are various other factors that
need to be considered. One obvious issue is the overall cost of the system. This
significantly influences design decisions as the detectors represent a significant
fraction of the overall production cost. The choice of detector material, the
thickness of the detectors, the diameter of the detector ring and the axial extent
of the detectors all contribute to the total cost of the system. Another important
design issue that affects the overall cost is integration of the PET system with a
second modality in a combined scanner. In most cases, this means integration
of the PET subsystem with a CT subsystem, although combined PET/magnetic
resonance (MR) scanners exist and present more significant technical challenges.
Combined PET/CT scanners have effectively replaced stand-alone PET for
clinical applications and the optimal CT configuration included in a combined
system is limited mainly by cost concerns. In practice, this means the number of
slices to include in the multi-detector CT system. High performance multi-detector
CT is important for systems intended for cardiac PET/CT applications, whereas
lower slice capability may be adequate for oncological PET/CT. Other relevant
design issues are related to the computer workstation used for acquiring and
processing data. These include fast image reconstruction times, convenient
integrated control of both PET and CT subsystems, and seamless integration with
other institutional information technology systems such as clinical information
systems and image archive systems.
361
CHAPTER 11
11.3.3. Detector systems

11.3.3.1. Radiation detectors
Although different radiation detector designs have been used in PET,
almost all current systems adopt an approach based on scintillation detectors.
Scintillation detectors are inorganic crystals that emit scintillation light in the
visible range when high energy photons are incident upon them (Fig.11.31). This
light is converted to an electrical signal by means of a photodetector, usually a
PMT, coupled to the crystal material. Various scintillator materials have been
used in PET, including thallium doped sodium iodide (NaI(Tl)), bismuth
germanate (BGO) and cerium doped lutetium oxyorthosilicate (LSO). Table11.1
shows some of the properties of the crystal materials that are relevant for PET
applications. The properties of an ideal crystal for PET would include a high
stopping power for 511keV photons (high linear attenuation coefficient); short
scintillation light decay time to reduce dead time and allow short coincidence
time windows to reduce random coincidences; and high light output. High light
output enables good energy resolution, which gives rise to improved scatter
FIG.11.31. Example of bismuth germanate crystals used for PET (a). Bismuth germanate
samples photographed under room lighting (b) and in the presence of X ray irradiation
and dimmed room lighting (c). The scintillation light seen in (c)is due to the interaction of
radiation with the crystals, which causes electrons to become excited. When they return to
their ground state, energy is emitted, partly in the form of visible light.
362
rejection. It also affords cost savings in the construction of a complete scanner

system as the number of photodetectors required to resolve a given number of
crystal elements can potentially be reduced.
TABLE11.1. PROPERTIES OF SOME OF THE SCINTILLATORS USED
IN PET. LINEAR ATTENUATION COEFFICIENTS AND ENERGY
RESOLUTION ARE QUOTED FOR 511keV
Property
NaI
BGO
LSO
Linear attenuation coefficient (cm )
0.34
0.95
0.87
Scintillation decay constant (ns)
230
300
40
100%
15%
75%
6.6
10.2
10.0
Relative light output

Energy resolution (%)
Note: BGO: bismuth germanate; LSO: lutetium oxyorthosilicate.
Although NaI(Tl) is ideal for lower energy single photon imaging, its
relatively low linear attenuation coefficient for 511keV photons makes it less
attractive for PET applications. Sensitivity could potentially be increased by
increasing the thickness of the crystals, which are typically 13cm thick.
However, the scope for substantially increasing crystal thickness is limited as it
results in a loss of spatial resolution. This is because thicker crystals are prone
to more significant depth of interaction problems as the apparent width of the
detector increases for sources located off-centre. Thin crystals composed of a
material with a high stopping power for 511keV photons are, thus, desirable
to ensure best possible sensitivity while maintaining spatial resolution. For this
reason, BGO and, more recently, LSO have replaced NaI(Tl) as the scintillator of
choice for PET.
BGO has the advantage of a high stopping power for 511keV photons and
has become an important scintillator for PET applications. However, it is not ideal
in many respects as it has relatively poor energy resolution and a long crystal decay
time. The poor energy resolution translates into a limited ability to reject scatter
via energy discrimination. In addition, the long decay time translates into a greater
dead time and increased number of random coincidences at high count rates. As
such, BGO is well suited for scanner designs thatminimize scatter and count rate
via physical collimation, such as those with interplane septa (seeSection 11.3.4.2).
Attempts to increase sensitivity by removing the interplane septa typically result
in a high scatter, high count rate environment for which BGO is not ideal.
Although LSO has a lower linear attenuation coefficient than BGO, its
shorter crystal decay time and slightly improved energy resolution convey
363
CHAPTER 11
significant advantages as a PET scintillator. LSO has become the scintillator of

choice for scanner designs that operate without interplane septa because its short
decay time makes it well suited for high count rate applications. The fast decay
time of LSO also enables a time of flight (TOF) data acquisition mode that will be
discussed in Section 11.3.4.4. LSO has proved to be a successful crystal for PET
detector applications despite the fact that the material contains around 2.6% 176Lu,
which is itself radioactive. A component of the emissions from 176Lu is detected
within the energy acceptance window and the dominant effect is to contribute
random coincidences. In practice, the increased randoms rate is not a major
problem for clinical studies, and the naturally occurring radiation has even been
used for quality assurance. Lutetium-176 has a half-life of 3.81010 a and, thus,
provides a long lived source of radiation that can be used to check consistency of
detector response without the need for external sources. It should be noted that, for
commercial reasons, some PET systems employ cerium doped lutetium yttrium
oxyorthosilicate (LYSO(Ce)) which has substantially similar properties to LSO.
11.3.3.2. Detector arrangements
As discussed above, the interaction of 511keV annihilation radiation
with the scintillation crystals gives rise to optical light that can be detected by
a suitable photodetector. A photodetector is a device that produces an electrical
signal when stimulated by light of the sort emitted by a scintillation detector.
For most PET applications, PMTs have been the preferred photodetector because
their high gain results in an electrical output with a good signal to noise ratio.
In addition, PMT output is proportional to the intensity of the incident light
and, thus, proportional to the energy deposited in the crystal. This provides a
mechanism for selective acceptance of only those detection events with energies
within a specific range and can be used to reject scattered photons. In addition,
PMTs provide high amplification with little degradation in the timing information
that is essential for electronic collimation. Although PMTs are by far the most
widely used photodetector for PET applications, they are somewhat bulky and
highly sensitive to magnetic fields. For these reasons, they are often not used in
combined PET/MR systems where space is limited and operation in high magnetic
fields is a requirement. In these and some other applications, semiconductor
based photodiodes are an alternative to PMTs. It should be noted that in these
applications, the semiconductor device is used in conjunction with a scintillation
detector and is used to detect scintillation light, not the annihilation photons.
Avalanche photodiodes have much lower gains than PMTs but can be very small
and have been shown to be effective in high magnetic field environments. Their
low gain requires very low noise electronics and they are also sensitive to small
temperature variations.
364
Space and cost constraints mean that individual scintillation crystals are
not usually coupled directly to individual photodetectors in a one to one fashion.
Instead, the most common arrangement is a block detector in which a group of
crystal elements share a smaller number of PMTs (Fig.11.32). The design of each
block varies between manufacturers and scanner models but usually involves a
matrix of crystal elements, a light guide and four PMTs. An example configuration
might be an 88 array of closely packed 4.4mm4.0mm30mm crystal
elements, where the longest dimension is in the radial direction to maximize
detection efficiency. The light guide allows light to be shared between four
circular PMTs and the relative light distribution depends on the location of the
crystal in which the photon interacted. The (x, y) position of the detection event is
calculated from the outputs of the four PMTs using a weighted centroid algorithm,
similar to the Anger logic of a gamma camera. Although individual crystals can
be identified in this way, the response is not linear throughout the block due to
differences in the locations of the different crystal elements relative to the PMTs.
Experimentally determined look-up tables are used to relate the measured (x, y)
position to a corresponding detector element, effectively performing a form of
linearity correction. In this way, only four PMTs are needed to localize signals
from a much greater number of crystal elements. The number of crystal elements
divided by the number of PMTs in a PET system has been referred to as the
encoding ratio. A high encoding ratio implies lower production costs and is,
therefore, desirable.
FIG.11.32. (a)A PET detector block consisting of scintillator material coupled to an array of
photomultiplier tubes. The scintillator is cut into an array of individual crystal elements. Four
photomultiplier tubes are used to read out the signal from the 88 array of crystal elements.
(b)The x and y position of each photon is determined from the signal measured by each of the
four photomultiplier tubes labelled AD, using the equations shown.
365
CHAPTER 11
One of the advantages of the design described above is that each block
operates independently of its surrounding blocks. This leads to good count rate
performance as light is not transferred between blocks and the PMTs of one
block are unaffected by detection events in an adjacent block. An alternative
arrangement, referred to as quadrant sharing, increases the encoding ratio by
locating the PMTs at the corners of adjacent blocks. This arrangement differs
from the conventional block design in that each PMT can now be exposed to
light from up to four different blocks. This can result in better spatial resolution
and a higher encoding ratio but is also susceptible to greater dead time problems
at high count rates.
Another alternative to the block design adopts an approach similar to
that used in conventional gamma cameras. These Anger-logic designs involve
detector modules that have a much larger surface area compared to conventional
block detectors, e.g. 92mm176mm. Each module is comprised of many small
crystal elements which are coupled, via a light guide, to an array of multiple
PMTs. Light is spread over a larger area than in the block design and positional
information is obtained using Anger-logic in the same way as a gamma camera.
The PMTs used in this design are typically larger than those used in block
detectors, increasing the encoding ratio. The larger area detector modules
encourage more uniform light collection compared to block designs, which leads
to more uniform energy resolution. However, a disadvantage of this design is that
the broad light spread among many PMTs can lead to dead time problems at high
count rates.
11.3.3.3. Scanner configurations
The detectors described above form the building blocks used to construct
complete scanner systems. Various scanner configurations have been developed,
although the dominant design consists of a ring of detectors that completely
surrounds the patient (or research subject) in one plane (Fig.11.33(a)). As with
other scanner systems, this plane is referred to as the transverse or transaxial
plane and the direction perpendicular to this plane is referred to as the axial or
z direction. Several rings of detectors are arranged in a cylindrical geometry,
allowing multiple transverse slices to be simultaneously acquired. As coincidence
detection requires two opposing detectors, a full ring system of this sort allows
coincidence data to be acquired at all angles around 180. Although complete
angular coverage is achieved in the transverse plane, there is much more limited
coverage in the axial direction. Cost constraints and, to some extent, limited
patient tolerance of extended tunnels mean that the detector rings usually extend
for only a few centimetres in the axial direction. Human whole body systems
typically have an AFOV of around 1520cm, although the trend in scanner
366
design has been to increase the AFOV, thus increasing both sensitivity and the
number of transverse slices that can be simultaneously acquired.
FIG.11.33. (a) Full ring PET system shown in the transverse plane, indicating how each
detector can form coincidence events with a specific number of detectors on the opposite side
of the ring. For clarity, this fan-like arrangement of lines of response is shown for only eight
detectors. The dashed line indicates how the imaging field of view is necessarily smaller than
the detector ring diameter. (b) PET system shown in side elevation, indicating the limited
detector coverage in the z direction. The shaded area indicates the coincidence field of view.
The dashed lines indicate the singles field of view. End shields reduce, but do not eliminate,
detection of single photons from outside of the coincidence field of view when operating in
3D mode.
The diameter of the detector ring varies considerably between designs and
this dimension reflects the intended research or clinical application. Small animal
systems may have ring diameters of around 15cm, brain oriented human systems
around 47cm and whole body human systems around 90cm. Systems with ring
diameters that can accommodate the whole body are clearly more flexible in terms
of the range of studies that can be performed. However, smaller ring diameters
have advantages in terms of increased sensitivity, owing to a greater solid angle
of acceptance, and potentially better spatial resolution, owing to reduced photon
non-collinearity effects. It should be noted that the spatial resolution advantage is
complicated by greater depth of interaction problems as the detector ring diameter
decreases and shorter crystals or depth of interaction measurement capability
may be required. Furthermore, the effective imaging FOV is always smaller than
the detector ring diameter because the acquisition of coincidence events between
all possible detector pairs (such as those between nearby detectors in the ring)
367
CHAPTER 11
is not supported. In addition, PET systems have annular shields at the two ends
of the detector ring that reduce the size of the patient port. These end shields
are intended to decrease the contribution of single photons from outside the
coincidence FOV (Fig.11.33(b)). The coincidence FOV refers to the volume that
the detector system surrounds, within which coincidence detection is possible.
Single photons originating from outside the coincidence FOV cannot give rise to
true coincidence events but may be recorded as randoms and can also contribute
to detector dead time. Reducing the size of these end shields allows the patient
port size to be increased but also leads to greater single photon contamination.
Unlike rotating camera SPECT systems, where different projections are
acquired in a sequential fashion, full ring PET systems simultaneously acquire
all projections required for tomographic image formation. This has an obvious
advantage in terms of sensitivity, and it also enables short acquisition times,
which can be important for dynamic studies. Full ring systems are, however,
associated with high production costs and, for this reason, some early PET
designs employed a partial ring approach. In these designs, two large area
detectors were mounted on opposite sides of the patient and complete angular
sampling was achieved by rotating the detectors around the z axis. Gaps in the
detector ring led to reduced sensitivity and the partial ring design is now usually
reserved for prototype systems. Another related approach to PET system design
was to use dual head gamma cameras modified to operate in coincidence mode.
The use of modified gamma cameras allowed for lower cost systems capable of
both PET and SPECT. However, the poor performance of NaI based PET means
that this approach has now been discontinued.
Current clinical systems have an AFOV that is adequate to cover most
individual organs but in order to achieve coverage of the whole body, patient
translation is required. Given the clinical importance of whole body oncology
studies, the mechanism for translating the patient through the scanner has
become an important component of modern PET systems. The patient bed or
patient handling system has to be made of a low attenuation material but must
still be able to support potentially very heavy patients. It must be capable of a
long travel range, so as to allow whole body studies in a single pass without the
need for patient repositioning. Precise motion control is also critical, particularly
for PET/CT systems where accurate alignment of the two separately acquired
modalities is essential. Advanced patient handling systems have been specifically
developed for PET/CT to ensure that any deflection of the bed is identical for
both the CT and PET acquisitions, thus ensuring accurate alignment irrespective
of patient weight. Although most patient beds have a curved shape for improved
patient comfort and better mechanical support, many manufacturers can also
provide a flat pallet that is more compatible with radiation treatment positioning.
368
11.3.4. Data acquisition

11.3.4.1. Coincidence processing
The basis of coincidence detection is that pairs of related 511keV
annihilation photons can be associated together by the detector system based
upon their times of measurement. Two photons detected within a short time
interval are assumed to have arisen from the same positronelectron annihilation
and a coincidence event is recorded. The time interval determining when events
are considered to be coincident is denoted 2 and is a system parameter that is not
usually adjustable by the user. In order tominimize random coincidences, this
interval should be kept as short as possible and for typical BGO based systems,
2 may be around 12 ns. Shorter time windows are made possible by detector
materials such as LSO that have faster scintillation decay times and, thus, better
time resolution. Further reductions in the coincidence window are limited by
differences in the arrival times of the two photons. For an electronpositron
annihilation taking place at the edge of the transverse FOV, one photon would
have to travel only a short distance, whereas the other photon might have to
travel almost the diameter of the detector ring. Assuming a 90cm ring diameter
and a speed of light of 3108 m/s, it can be seen that a maximum time difference
of around 3 ns can be expected.
When a photon is incident upon a PET detector, an electrical pulse is
generated (Fig.11.34). A constant fraction discriminator then produces a digital
logic pulse when the detector signal reaches a fixed fraction of the peak pulse
height. This digital logic pulse is defined to have a duration and is fed into
a coincidence module that determines whether it forms a coincidence event
with any of the signals from other detectors in the ring. A coincidence event
is indicated if there is an overlap in time between separate logic pulses from
two different detectors. In other words, a coincidence would be identified if
two detection events were recorded within a time interval no greater than . A
photon detected at time t can, thus, form a coincidence with another photon
detected within the interval t + . It can alternatively form a coincidence with
an earlier photon detected within t . For this reason, 2 is often referred to as
the coincidence time window. Consistent timing of signals from every detector
in the system is clearly essential to ensure that true coincidences are effectively
captured within the coincidence time window. Differences in performance of
the various detector components and cable lengths can introduce variable time
offsets. For this reason, time alignment corrections are performed to characterize
and compensate for timing differences between different detectors in the ring.
369
CHAPTER 11
Amplifier
Pulse
height
analyser
To sorter
Coincidence
processor
Amplifier
Pulse
height
analyser
FIG.11.34. Coincidence circuits allow two photon detection events to be associated with each
other based upon their arrival times. Photons detected at A and B produce signals that are
amplified and analysed to determine whether they meet the energy acceptance criteria. Those
signals that fall within the energy acceptance window produce a logic pulse (width ) that is
passed to the coincidence processor. A coincidence event is indicated if both logic pulses fall
within a specified interval (2).
Coincidence detection assumes that only two photons were detected, but
with multiple disintegrations occurring concurrently, it is possible for three or
more photons to be recorded by separate detectors within the coincidence time
window. When this occurs, it is unclear which pair of detectors corresponds to a
legitimate coincidence and multiple events of this sort are often discarded. This
circumstance is most likely to occur when there is a large amount of activity in
or around the FOV, and it contributes to count loss at high count rates. Another
possible scenario is that only one photon is detected within the coincidence time
window and no coincidence event will be recorded. These single photon detection
events are a result of a number of reasons: the angle of photon emission was such
that only one of the two annihilation photons was incident upon the detectors;
one of the two annihilation photons was scattered out of the FOV; one of the two
annihilation photons was absorbed within the body; and photons originating from
outside the coincidence FOV. Although these single photons cannot form true
coincidences, they are a major source of randoms.
The mechanism described above records what are known as prompt
coincidence events which consist of true, random and scattered coincidences.
370
The relative proportion of each component depends on factors such as the count
rate, the size of the attenuation distribution (patient size) and the acquisition
geometry (2D or 3D). Only the trues component contributes useful information
and the randoms and scattered coincidences need to beminimized. The randoms
component is maintained as low as possible by setting the coincidence time
window to the shortest duration consistent with the time resolution of the system.
The scatter component is maintained as low as possible by energy discrimination.
In addition to providing positional and timing information, the pulse produced
by the detectors can be integrated over time to provide a measure of the energy
deposited in the detector. In a block detector with four PMTs, the sum of the
signals from each PMT is proportional to the total amount of scintillation light
produced and, thus, the total energy deposited in the detector material. Under
the assumption that the photon was completely absorbed in the detector, this
signal provides a measure of the photons energy and can be used to reject lower
energy photons that have undergone Compton scattering within the patient. In
practice, the energy resolution of most PET detector systems is such that the
energy acceptance window must be set quite wide to avoid rejecting too many
unscattered 511keV photons. For BGO based systems, an energy acceptance
range of 350650keV is typical. As small-angle scatter can result in only a small
loss of energy, many of these scattered photons will be accepted within the energy
window, despite the fact that they do not contribute useful information. Energy
discrimination, therefore, reduces, but does not eliminate, scattered coincidence
events and additional compensation is required.
11.3.4.2. Data acquisition geometries
Scanners consisting of multiple detector rings provide extended axial
coverage and are advantageous for rapid acquisition of volumetric data.
However, the presence of multiple detector rings raises issues concerning the
optimum combinations of detectors that should be used to measure coincidence
events. In a system with only one ring of detectors, the acquisition geometry is
simple as each detector measures coincidence events with other detectors on
the opposite side of the same ring. When additional detector rings are added to
the system, it is possible to allow coincidence events to be recorded between
detectors in different rings. This means including photons that were emitted in
directions that are oblique to the transverse plane. Given that photons are emitted
in all directions, increasing the maximum ring difference increases the angle of
obliqueness that is accepted and, therefore, increases system sensitivity. The data
acquisition geometry refers to the arrangement of detector pairs that are permitted
to form coincidence events and, in practice, involves the presence or absence
of interplane septa (Fig.11.35). Data acquisition with septa in place is referred
371
CHAPTER 11
to as 2D mode; data acquisition without any interplane septa is referred to as

3D mode. The 2D/3D designation refers to the acquisition geometry rather
than the resulting images as both modes produce similar volumetric images.
In 2D acquisition mode, an array of septa is inserted between the detector
rings. These septa are annular and are typically made of tungsten. The purpose of
the septa is to physically absorb photons incident at large oblique angles relative
to the transverse plane, allowing only those photons incident approximately
orthogonal to the z axis of the scanner. These septa differ significantly from
gamma camera parallel-hole collimators as in the PET case, no collimation is
provided within the transverse planes. By physically rejecting almost all oblique
photons from reaching the detectors, the count rate is substantially reduced,
resulting in a low rate of random coincidences and low detector dead time. In
addition, 2D acquisition is associated with a low rate of scattered coincidence
events since only photons emitted in and scattering within a transverse plane can
pass through the septa. However, if a Compton interaction occurs, the likelihood
is that the scattered photon will emerge at an oblique angle. Photons that undergo
a Compton interaction are scattered through an angle that is distributed over 4,
so although the photon may be scattered in-plane, out-of-plane scatter is more
likely. Photons scattered through oblique angles will be absorbed by the septa,
effectively reducing the fraction of scattered coincidences that are measured.
FIG.11.35. (a)2D and (b)3D acquisition geometries. In 2D mode, a series of annular septa
are inserted in front of the detectors so as to absorb photons incident at oblique angles. In 3D
mode, these septa are removed, allowing oblique photons to reach the detectors. 3D mode is
associated with high sensitivity but also increased scatter and randoms fractions, the latter
partly due to single photons from outside the coincidence field of view.
372
In 3D acquisition mode, the septa are entirely removed from the FOV and
there is no longer any physical collimation restricting the photons that are incident
upon the detectors. Coincidence events can be recorded between detectors in
different rings and potentially between all possible ring combinations. Photons
emitted at oblique angles with respect to the transverse plane are no longer
prevented from reaching the detectors, and system sensitivity is substantially
increased compared to 2D acquisition. Sensitivity gains by a factor of around
five are typical, although the exact value depends on the scanner configuration
and the source distribution. In 2D mode, sensitivity varies slightly between
adjacent slices but does not change greatly over the AFOV. In 3D mode, the
sensitivity variation in the axial direction is much greater and has a triangular
profile with a peak at the central slice. The triangular axial sensitivity profile
can be understood by considering a point source centrally located in the first
slice at one of the extreme ends of the scanner. True coincidence events can only
be recorded between detectors in the first ring. As the source is moved towards
the central slice, coincidence events can be recorded between an increasing
number of detector ring combinations, leading to an increase in sensitivity. As a
consequence of the substantial sensitivity increase, 3D acquisition is associated
with higher detector count rates, leading to more randoms and greater dead time
than corresponding acquisitions in 2D mode. Furthermore, 3D mode cannot
take advantage of the scatter rejection afforded by interplane septa and, as a
result, records a greatly increased proportion of scattered coincidence events.
The advantage of 3D acquisition is its large increase in sensitivity
compared to 2D acquisition. This would be expected to result in images with
improved statistical quality or, alternatively, comparable image quality with
shorter scan times or reduced administered activity. In practice, evaluating
the relative advantage of 3D acquisition is complex as it is associated with
substantial increases in both the randoms and scatter components. Both of these
unwanted effects can be corrected using software techniques, but these corrections
can themselves be noisy and potentially inaccurate. Furthermore, the relative
contribution of randoms and scattered photons is patient specific. The randoms
and scatter fractions are defined as the randoms or scatter count rate divided by
the trues rate, and both increase with increasing patient size. Both randoms and
scatter fractions are substantially higher in 3D compared to 2D mode. In 3D
mode, scatter fractions over 50% are common, whereas 15% is more typical for
2D mode. Randoms fractions are more variable as they depend on the study, but
randoms often exceed trues in 3D mode.
A figure of merit that is sometimes useful when considering the
performance of scanner systems is the noise equivalent count rate (NECR). The
NECR is equivalent to the coincidence count rate that would have the same noise
properties as the measured trues rate after correcting for randoms and scatter.
373
CHAPTER 11
NECR is commonly used to characterize 3D performance and, since the relative

proportion of the different kinds of coincidence events is strongly dependent on
object size, standardized phantoms have been developed. It is computed using:
NECR =
T2
(11.12)
T + S + 2 fR
where
T, S and R are the true, scatter and random coincidence count rates, respectively;
and f is the fraction of the sinogram width that intersects the phantom.
For a given phantom, the NECR is a function of the activity in the FOV
and is usually determined over a wide activity range as a radioactive phantom
decays (Fig.11.36). The reason for this count rate dependence is twofold: the
randoms rate increases as the square of the single photon count rate (which is
approximately proportional to the activity in the FOV) and the sensitivity of the
scanner for trues decreases with increasing count rates as detector dead time
becomes more significant.
An important factor when considering the relative performance of 2D and
3D acquisition modes is the characteristics of the detector material. In 2D mode,
the septa substantially reduce dead time, randoms and scatter, making the poor
timing and energy resolution of BGO less of a limitation. BGO based systems
are, thus, well suited to 2D acquisition mode. However, for BGO, the sensitivity
advantage of 3D acquisition mode is substantially offset by the high randoms
and scatter fractions that are encountered. For systems based on detectors such as
LSO, the improved timing resolution can be used to reduce the coincidence time
window and, thus, reduce the randoms fraction. The improved energy resolution
also allows the lower level energy discriminator to be raised, resulting in a lower
scatter fraction. LSO or similar fast detector materials are, thus, well suited to
3D acquisition mode. The introduction of these detectors, along with improved
reconstruction algorithms for 3D data, means that 3D acquisition mode now
dominates. Many scanner systems no longer support 2D mode as this allows the
septa to be completely removed from the design, reducing cost and potentially
increasing the patient port diameter.
374
Count rate (kcps)
Activity concentration (kBq/mL)
FIG.11.36. The relative proportion of true, random and scattered coincidence events as
a function of activity in the field of view. At low activities, the true coincidence count rate
increases linearly with activity. However, at higher activities, detector dead time becomes
increasingly significant. The trues rate increases less rapidly with increasing activity and
can even decrease at very high activities. The randoms count rate increases with increasing
activity as a greater number of photons are detected. The scatter count rate is assumed to be
proportional to the trues rate. Scanner count rate performance can be characterized using
the noise equivalent count rate (NECR), which is a function of the true, random and scatter
coincidence count rates.
11.3.4.3. Data organization

The data recorded during a conventional PET acquisition are the total
number of coincidence events measured between the various detector pairs. These
data are typically binned into 2D matrixes known as sinograms (Fig. 11.37). If
a 2D acquisition geometry is considered, each row of the sinogram represents a
projection of the radionuclide distribution at a particular angle around the patient.
These projections consist of coincidence events recorded between pairs of
detectors, where each detector pair forms LORs that are approximately parallel to
each other. The number of counts in each element of the projection is proportional
to a line integral of the radionuclide distribution within the limitations imposed
by the various physical effects such as scatter and attenuation. The sinogram is
375
CHAPTER 11
indexed along the y axis by angle and the x axis by distance. For a full ring system,
angular sampling is usually evenly spaced over 180 but the sampling along each
row is slightly non-linear. The separation of adjacent elements in the projection
decreases towards the edges of the FOV owing to the ring geometry. Correction
for this effect, known as arc correction, is required and is usually implemented
during image reconstruction. Adjacent elements within a particular sinogram row
would be expected to be associated with two parallel LORs joining detector pairs
that are next to each other in the ring. In practice, improved sampling is achieved
by also considering LORs that are offset by one detector. Despite the fact that
these LORs are not exactly parallel to the others, these data are inserted into
the sinogram rows as if they came from virtual detectors positioned in the gaps
between the real detectors.
(a)
(b)
s
s
FIG.11.37. Full ring PET scanners simultaneously measure multiple projections at different
angles with respect to the patient. An example showing the orientation of two parallel
projections is shown in (a). Projection data of this sort are typically stored in sinograms;
an example is shown in (b). In a sinogram, each row represents a projection at a different
angle . Each projection is made up of discrete elements that are indexed by s and contain the
number of coincidence counts recorded along individual lines of response. The two example
projections shown in (a)are also highlighted in sinogram (b).
The prior discussion of 2D acquisition mode only considered coincidence

events between detectors in a single ring, referred to as a direct plane. In practice,
the interplane septa do not completely eliminate the possibility of coincidence
events being detected between different nearby rings (Fig.11.38). Inclusion
of these slightly oblique events is advantageous as it increases sensitivity.
376
Coincidence events between detectors in immediately adjacent rings are combined

into a sinogram that is considered to have been measured in a plane located
between the two detector rings. This plane is referred to as a cross plane and is
considered to be parallel to the direct planes, despite the fact that the contributing
LORs are slightly oblique to these planes. As well as increasing sensitivity,
inclusion of these cross planes increases axial sampling by producing 2N 1
slices from an N ring scanner. Sensitivity is further increased for cross planes by
extending the ring difference between which coincidences are allowed from one
to three or higher odd numbers. This principle is also applied to direct planes,
resulting in coincidence events not just within the same ring but also between
detectors with ring differences of two or higher even numbers. It should be noted
that the data obtained from these different ring combinations are added together,
so individual sinograms actually consist of LORs that were measured at slightly
different oblique angles. The total number of ring combinations contributing to
a direct plane plus those contributing to a cross plane is sometimes referred to as
span. Within the limitations imposed by the septa, span can be increased, resulting
in increased sensitivity and degraded spatial resolution in the axial direction.
Detector rings
Septa
Lines of
response
FIG.11.38. Side elevation of an eight ring PET scanner in 2D ((a) and (b)) and
3D (c)acquisition modes. (a)Lines of response joining opposing detectors in the same ring
forming direct planes. (b)Lines of response between detectors in adjacent rings. These lines
of response are averaged to form cross planes (dotted line) that are assumed to be located at
the mid-point between adjacent detectors. Both direct and cross planes are simultaneously
acquired during 2D acquisition. (c)3D acquisition in which each ring is permitted to form
coincidence events with all other rings.
377
CHAPTER 11
When the septa are removed, as is the case in 3D acquisition mode, there is
no longer any physical restriction on the detector rings that can be used to measure
coincidence events. An N ring scanner could have a maximum ring difference of
N 1, resulting in up to N 2 possible sinograms. In 2D mode, such a system
would have a total of 2N 1 sinograms, so it can be seen that the total volume
of data is substantially higher in 3D mode. In order to reduce this volume for
ease of manipulation, the maximum ring difference can be reduced. This has the
effect of introducing a plateau on the axial sensitivity profile, converting it from a
triangular to a trapezoidal form. Additionally, several possible ring combinations
can be combined in a similar fashion to that indicated for 2D acquisition mode.
It should be noted that 3D acquisition mode results in data that are redundant
in the sense that only a subset of the sinograms (those in the transverse planes)
are required for tomographic image reconstruction. The purpose of acquiring the
additional oblique data is to increase sensitivity and reduce statistical noise in the
resulting images.
In addition to the sinogram representation described above, some scanners
also support list-mode acquisition. In this mode, coincidence events are not
arranged into sinograms in real time but are recorded as a series of individual
events. This stream of coincidence events is interspersed with time signals and
potentially other signals from ECG or respiratory gating devices. These data can
be used as the input to list-mode image reconstruction algorithms but may also be
sorted into sinograms prior to image reconstruction. The advantage of acquiring
in list-mode is that the sorting of the data into sinograms can be performed
retrospectively. This provides a degree of flexibility that is very helpful when data
are acquired in conjunction with physiological gating devices or when sequential
images over time are of interest. For example, separate ECG gated and dynamic
time series images can be obtained from the same cardiac list-mode acquisition.
Furthermore, certain parameters can be retrospectively adjusted and do not have
to match the parameters chosen at the time of acquisition.
11.3.4.4. Time of flight
Detectors operating in coincidence mode provide spatial information related
to individual positronelectron annihilations but this information is not sufficient
to determine the exact location of each event. A line joining the two detectors
can be assumed to intersect the site of the annihilation but the exact position
along this line cannot be determined. For this reason, PET systems measure
signals from multiple events and the resulting projections are used to reconstruct
images using computed tomography. However, it has long been appreciated that
the difference in the detection times of the two annihilation photons provides
a mechanism for precisely localizing the site of individual positronelectron
378
annihilations (Fig.11.39). Given that photons travel at the speed of light,

essentially irrespective of the composition of the material through which they
pass, the difference in the arrival times of the two photons can potentially be
used to localize their original point of emission. This is clearly attractive because
it means that each coincidence measurement provides significantly more
information, promising substantial improvements in image statistical quality.
FIG.11.39. (a) A coincidence event detected along a line of response between

detectors A and B. The average time difference between the two detectors is given by
(x + x)/c (x x)/c=2x/c, where c is the speed of light. (b)With conventional PET, no
information is available about the location of the annihilation event along the line of response.
During reconstruction, the event is assigned with equal weight to all pixels between A and B.
(c)With time of flight PET, the time difference between the signals recorded at detectors A and
B is used to estimate the position of the annihilation event along the line of response. During
reconstruction, events are weighted according to the detector time difference and a function
that reflects the limited time resolution of the system.
Incorporating information derived from differences in the photon arrival

times has been referred to as TOF mode and a number of PET systems have
been developed that exploit this approach. A prerequisite for TOF PET systems
is high timing resolution (t) as this determines the spatial uncertainty (x)
with which the annihilation event can be localized. The two parameters are
related by x=ct/2, where c is the speed of light (3108 m/s). According
to this equation, a timing resolution of 66 ps is required to achieve 1cm depth
resolution. Early TOF PET systems used detector materials that, although they
had good timing resolution, suffered from a poor stopping power for 511keV
379
CHAPTER 11
photons. The resulting low sensitivity of these devices could not be offset by the
improved signal to noise ratio provided by the TOF information and interest in
the method declined. Interest was subsequently rekindled with the introduction
of LSO based systems, which have been able to combine timing resolutions
of around 600 ps with high sensitivity. A timing resolution of 600 ps translates
to a spatial uncertainty of 9cm which, although clearly worse than the spatial
resolution that can be achieved with conventional PET, does represent useful
additional information.
In addition to the high performance required for conventional PET, TOF
PET requires scanners optimized for high timing resolution. The additional TOF
information has data management considerations because an extra dimension has
been added to the dataset. TOF data may be acquired in list-mode and fed directly
to a list-mode reconstruction algorithm that is optimized for TOF. Alternatively,
the data may be reorganized into sinograms where the sinograms have an
additional dimension reflecting a discrete number of time bins. Each coincidence
event is assigned to a particular sinogram depending on the difference in
the arrival times of the two photons. TOF sinograms also require dedicated
reconstruction algorithms that incorporate the TOF information into the image
reconstruction. An interesting feature of TOF PET is that the signal to noise ratio
gain provided by the TOF information is greater for larger diameter distributions
of radioactivity. This is related to the fact that the spatial uncertainty x becomes
relatively less significant as the diameter increases. This has potential benefits for
body imaging, particularly in large patients where high attenuation and scatter
mean that image quality is usually poorest.
11.3.5. Data corrections
11.3.5.1. Normalization
Normalization refers to a software correction that is applied to the measured
projection data in order to compensate for variations in the sensitivity of
different LORs. Without such a correction, images display systematic variations
in uniformity and pronounced artefacts that include spike and ring artefacts at
the centre of the FOV (Fig.11.40). It is somewhat analogous to the uniformity
correction applied to gamma camera images. Sources of sensitivity variations
include:
Detector efficiency variations: The detection efficiency of a particular
LOR depends on the efficiencies of the individual detectors involved.
Individual detectors can have variable efficiency due to differences in PMT
380
gain, differences in the performance of individual crystal elements and the

position of the detector element within the larger detector array.
Geometric effects: Geometric issues also influence the sensitivity of
different LORs. Individual detector elements contribute to multiple LORs
and measure photons that are incident over a range of angles. Photons
incident normal to the face of the detector will have a shorter thickness of
detector material in their path compared to those incident at more oblique
angles. This results in a greater probability of detection for photons incident
at more oblique angles. However, another geometric effect that exists in
full ring block based systems occurs towards the edges of each projection
where LORs are formed by detectors at larger oblique angles. Although
oblique angles lead to a greater thickness of crystal, they also decrease
the solid angle that the detectors present, reducing the sensitivity for those
particular LORs.
FIG.11.40. Sinograms corresponding to a centrally located uniform cylinder before

normalization (a) and after normalization (b). Transverse images are shown for reconstructions
without normalization (c)and with normalization (d). The artefacts and non-uniformity seen in
the image without normalization should be noted.
The acquisition geometry, either 2D or 3D mode, determines the way

different detector pairs are combined and, thus, also influences sensitivity
variations between different LORs. In addition, the presence of the interplane
septa has a shadowing effect that reduces sensitivity. For this reason, separate
381
CHAPTER 11
2D and 3D normalizations are required for systems capable of acquiring data in

both modes. Normalization files are experimentally determined correction factors
that are applied as multiplicative terms for each LOR. They are periodically
updated to reflect the current state of the detector system and are applied to all
subsequently acquired data. In order not to degrade the statistical quality of the
patient data, normalization coefficients need to be measured with low statistical
noise, and a variety of methods have been developed to achieve this.
11.3.5.2. Randoms correction
Randoms make up a potentially large component of all measured
coincidence events and, if left uncorrected, will contribute to a loss of image
contrast and quantitative accuracy. Random coincidences are generally smoothly
distributed across the FOV but the magnitude of the randoms component depends
on the count rate encountered during data acquisition. 3D acquisition mode
or studies involving large amounts of activity in or near the FOV are usually
associated with high randoms fractions. Randoms correction is essential for all
quantitative studies and is routinely implemented on almost all scanner systems.
One widely adopted correction method involves estimating the number
of randoms contributing to the prompts (trues + scatter + randoms) using an
additional coincidence circuit. This secondary coincidence circuit is acquired
simultaneously with the prompt measurement but is only sensitive to random
events. Preferential selection of randoms is achieved by delaying the logic pulse
from one of the detectors such that it cannot form a true coincidence event
between corresponding annihilation photons (Fig.11.41). Although the time delay
prevents measurement of true and scattered photons, it does not stop coincidence
events being recorded by chance between unrelated 511keV photons. As this
secondary, or delayed, coincidence circuit is identical to the prompt circuit in all
other respects, the number of counts in the delayed channel provides an estimate
of the randoms in the prompt channel. In many implementations, the delayed
data are automatically subtracted from the corresponding prompt data, providing
an on-line randoms correction. Alternatively, the prompt and delayed data can be
stored as separate sinograms for retrospective off-line subtraction, often as part
of a statistical image reconstruction algorithm.
The delayed channel does not identify and remove individual random
coincidences from the prompt measurement but, instead, estimates the average
number of randoms that might be expected. As the delayed channel records
counts in individual LORs over a limited scan duration, the randoms estimate is
often noisy, leading to increased statistical uncertainty in the corrected data after
randoms subtraction.
382
FIG.11.41. Diagram illustrating the concept of how a delayed coincidence circuit can be
used to estimate the number of random events in the prompt circuit. (a)Detection events from
two opposing detectors, indicating three coincidence events in the prompt circuit. (b) Data
from detector 2 delayed with respect to detector 1 and indicating one coincidence event in this
delayed circuit. The temporal delay prevents true coincidence events from being recorded in
the delayed circuit, but random coincidence events still occur with the same frequency as in
the prompt circuit. If data are acquired with sufficient statistical quality, the total number of
delayed coincidence events provides an estimate of the total number of randoms in the prompt
circuit.
An alternative method of randoms correction is to estimate the randoms

for each LOR using the singles rates at the corresponding detectors. If N1 and N2
are the singles rates at two opposing detectors, the rate of random coincidences
between these detectors is given by 2N1N2 where 2 is the coincidence timing
window. This approach has the advantage that singles count rates are substantially
higher than coincidence count rates and, therefore, lead to randoms estimates
with better statistical quality than the delayed channel method.
11.3.5.3. Attenuation correction
Despite their high energy, only a small fraction of the emitted 511keV
photon pairs escape the body without undergoing some form of interaction.
Compton interaction is the most likely mechanism and, depending on the energy
and direction of the scattered photon, may result in the detection of a scattered
coincidence event. However, it is more likely that the scattered photon will not
result in a coincidence event for a variety of reasons. The scattered photon may
emerge from the Compton interaction along a path that is not incident upon the
detectors. Alternatively, the scattered photon may undergo further Compton
383
CHAPTER 11
interactions, resulting in a lower energy and a greater likelihood of photoelectric

absorption. Even if the scattered photon does reach the detectors, it may have lost
so much energy that it does not meet the energy acceptance criteria of the scanner
and will be rejected. Thus, as well as creating scattered coincidence events,
Compton interactions lead to a much greater loss of true coincidence events. This
underestimation of true counts is referred to as attenuation.
One of the advantages of PET over SPECT is the ease with which
attenuation correction can be performed. This feature of coincidence detection can
be understood by considering a point source located at depth x within a uniformly
attenuating object with attenuation coefficient m. Figure11.42(a)shows an LOR
passing through the point source and intersecting a thickness D of attenuating
material. The probability of photon 1 escaping the object without undergoing any
interactions p1 is given by:
p1 =
I ( x)
= e x (11.13)
I (0)
where
I(x) is the beam intensity after passing through attenuating material of thickness
x;
and I(0) is the intensity in the absence of attenuation.
The probability that the corresponding photon 2 will also escape the object
is given by p2:
p2 =
I (D x)
= e ( Dx ) (11.14)
I (0)
The probability of both photons escaping the body such that a coincidence
event can occur is given by the product of p1 and p2:
p1 p 2 = e x e ( Dx ) = e D
(11.15)
It can be seen that the probability of a coincidence event occurring decreases

as the thickness of attenuating material increases. However, this probability is not
dependent on the location of the source along a particular LOR. This result differs
from the SPECT case where the attenuation experienced by a single photon
source along a particular LOR is strongly dependent on the distance between the
source location and the edge of the attenuating medium.
384
FIG.11.42. When considering both back to back photons along a particular line of response,
the attenuation experienced by a point source within the body (a)is independent of its location
along the line and is given by eD for a uniformly attenuating object. In (b), the positron
emitting transmission source is outside the patient but experiences the same attenuation as the
internal source when considering coincidence events along the same line of response.
From Eq.(11.15), it can be seen that photon attenuation reduces the

number of true coincidence events by a factor of emD. In practice, the body is not
composed of uniformly attenuating material, and the attenuation factor AF for a
particular LOR is given by:
( x ) dx
af = e
(11.16)
where m(x) refers to the spatially-variant distribution of linear attenuation

coefficients within the body and the integral is over the LOR joining opposing
detectors.
As this attenuation factor is dependent only on the material along the LOR,
it is identical to the attenuation experienced by an external source of 511keV
photons along the same line (Fig.11.42(b)). Consider an external source of
511keV photons that gives rise to a total of N0 coincidence events along a
particular LOR in the absence of any attenuation. When a patient is interposed,
the number of coincidence events falls to Nx where:
( x ) dx
N x = N 0e
(11.17)
From this equation, it can be seen that the attenuation factor can be obtained
by dividing Nx by N0. The attenuation correction factor is simply the reciprocal
of the attenuation factor and is given by N0/Nx. N0 can be obtained from what
385
CHAPTER 11
is referred to as a blank scan, as it is acquired with nothing in the FOV. Nx can

be obtained from what is referred to as a transmission scan, as it measures the
coincidence events corresponding to the photons that pass through the patient
while in the scanning position.
A variety of transmission systems have been developed for measuring
patient specific attenuation correction factors. Early transmission systems used
68
Ge (271 d half-life, decays to the positron emitter 68Ga) ring sources, although
poor scatter rejection meant that the ring configuration was superseded by rotating
rod sources. These sources, also consisting of 68Ge, were oriented parallel to the
z axis of the scanner and could be inserted close to the detectors at the periphery
of the PET FOV during a transmission acquisition. Once in position, they would
rotate around the patient in a similar fashion to the motion of an X ray tube in
CT. Coincidence events were recorded in 2D acquisition mode for all LORs
as the source rotated continuously around the patient. The patient transmission
data were used in conjunction with a similarly acquired blank scan to determine
an attenuation correction sinogram that was then applied to the patient emission
data, under the assumption that the patient did not move between scans. The
rotating rod configuration had the advantage that it enabled transmission data
to be acquired in conjunction with a spatial window that tracked the current
position of the rotating source. Coincidence events that were not collinear with
the current position of the source, such as scattered coincidences, could be
rejected, improving the quality of the transmission data. Rod windowing also
helped reduce contamination of the transmission measurement by coincidence
events not originating from the rotating sources but from a radiopharmaceutical
within the patient. The ability to acquire transmission data in the presence of a
positron emitting tracer within the body was of great practical significance as,
without this capability, lengthy protocols were required involving transmission
acquisition prior to tracer administration.
Rod windowing also provided the potential for simultaneous acquisition
of emission and transmission data, although cross-contamination meant that
separate emission and transmission acquisitions were usually preferred. The
disadvantage of acquiring emission and transmission data in a sequential fashion
was that scan times for both modes were necessarily lengthy in order to obtain
data with sufficient statistical quality. Increasing the rod source activity was not
an effective way of reducing transmission scan times as the source was located
close to the detectors and dead time at the near-side detectors quickly became
a limiting factor. As an alternative to 68Ge, single photon emitters such as 137Cs
(30 a half-life, 662keV) were used as a transmission source. Single photon
transmission sources had the advantage that they could be shielded from the
near-side detectors and could, thus, use much greater amounts of activity, leading
to data with improved statistical quality and, in practice, shorter scan times.
386
The fact that the single photon emissions were at 662keV as opposed to 511keV
and that the transmission data had a large scatter component was problematic but
could be effectively suppressed using software segmentation algorithms.
With the introduction of PET/CT, the need for radionuclide transmission
systems was eliminated as, with careful manipulation, the CT images can be
used not just for anatomic localization but also for attenuation correction. CT
based attenuation correction has a number of advantages, including the fact that
the resulting attenuation correction factors have very low noise due to the high
statistical quality of CT; rapid data acquisition, especially with high performance
multi-detector CT systems; insensitivity to radioactivity within the body; and no
requirement for periodic replacement of sources as is the case with 68Ge based
transmission systems. CT based attenuation correction is significantly different
from radionuclide based transmission methods because the data are acquired
on a separate, albeit well integrated, scanner system using X ray photons with
energies that are very different from the 511keV photons used in PET. Unlike
monoenergetic PET photons, the photons used in CT consist of a spectrum
of energies with a maximum value that is dependent on the peak X ray tube
voltage (kVp). CT Hounsfield units reflect tissue linear attenuation coefficients
that are higher than those applicable to PET photons as they are measured at an
effective CT energy (~70keV), which is substantially lower than 511keV. An
important step in the process of using CT images for attenuation correction is
to scale the CT images to linear attenuation coefficients that are applicable to
511keV photons. A number of slightly different approaches have been employed
but usually involve multi-linear scaling of the CT Hounsfield units using
functions specific for the X ray tubekVp setting (Fig.11.43). The methods used
are very similar to those described in Section 11.2.3.2 (Eq.(11.10)) for SPECT.
After scaling, the CT images are filtered, so as to have a spatial resolution that is
similar to that of the PET data and attenuation factors are calculated by integration
in a manner indicated by Eq.(11.16). The integration, or forward projection, is
performed over all directions measured by the PET system and, thus, provides
attenuation correction factors for all LORs.
CT based attenuation correction has proved to be very effective although
a number of potential problems require consideration. Patient motion,
commonly motion of the arms or head, can cause the CT and PET images to be
misregistered, leading to incorrect attenuation correction factors, which in turn
cause image artefacts and quantitative error. Respiratory motion can also lead to
similar problems as the CT and PET data are acquired over very different time
intervals. CT data acquisition is extremely short and usually captures a particular
phase in the respiratory cycle. In contrast, PET data are acquired over multiple
breathing cycles and the resulting images represent an average position that will
be somewhat blurred in regions where respiratory motion is significant.
387
CHAPTER 11
In the area around the lung boundary where there is a sharp discontinuity
in the bodys attenuation properties, respiratory motion can lead to localized
misregistration of the CT and PET images, and pronounced attenuation
correction artefacts. Another consideration for CT based attenuation correction
arises when the CT FOV is truncated such that parts of the body, usually the
arms, are not captured on the CT image or are only partially included. This
leads to under-correction for attenuation and corresponding artefacts in the PET
images.
m (511 keV) (cm1)
PET image
X ray tube voltage
CT image
CT (HU)
FIG.11.43. For PET attenuation correction, CT images have to be rescaled from Hounsfield
units (HU) to linear attenuation coefficients () appropriate for 511keV. Multi-linear scaling
functions have been used. The rescaled CT images are then forward projected to produce
attenuation correction factors that are applied to the PET emission data, prior to or during
image reconstruction.
11.3.5.4. Scatter correction

Scatter correction is required because the limited energy resolution of
PET systems means that scattered photons can only be partially rejected by
388
energy discrimination. Uncorrected scatter forms a background in reconstructed

images that reduces lesion contrast and degrades quantitative accuracy. This
scatter background is a complex function of both the emission and attenuation
distributions and is non-uniform across the FOV. In 2D mode, physical
collimation ensures that the scatter contribution is relatively low compared to
3D mode and approximate corrections, based on scatter deconvolution, have
been widely used. The form of the scatter distribution function can be measured
experimentally using line sources at different positions in a water phantom.
Analytical expressions derived from this scatter function can be determined and
convolved with the projection data from individual patient studies to estimate
the scatter distribution. This method assumes a uniform scattering medium and
has limited accuracy in areas such as the thorax. It also cannot account for scatter
between adjacent planes, which is significant for 3D acquisition mode.
An alternative algorithm that has been applied to 3D brain studies involves
a tail fitting approach. In brain studies, the LORs that do not pass through the
head are comprised of scatter that can be modelled by fitting a Gaussian function
to the tails of each projection. This function can be interpolated to the projections
that do pass through the head and used as an estimate of the scatter contribution
along these LORs. This method provides a first order correction for scatter and
has limited accuracy in areas of non-uniform attenuation or any area where the
tails of the projections cannot be accurately measured.
More accurate scatter correction can be achieved in 3D using a model
based approach. This method makes use of the physics of Compton scattering to
model the distribution of coincidence events for which one of the two photons
experienced a single scattering interaction. The assumption that the scatter
component in the measured data is dominated by these single scatter events has
been shown to be reasonable. An initial estimate of the radionuclide distribution
is first obtained from a reconstruction that does not include any scatter correction.
This image is then used, in conjunction with an attenuation map derived from CT
or transmission data, to estimate the scatter contribution to each LOR. Different
implementations have been developed, but each makes use of the KleinNishina
formula to determine the probability of a photon scattering through a certain
angle and being detected by a certain detector. Determining these probabilities
for all possible scattering positions and LORs is computationally demanding.
However, good scatter estimates can be obtained by interpolating data estimated
using a coarse grid of scattering locations and a subset of LORs. These data
can then be interpolated for all LORs, resulting in an estimate of the scatter
distribution that has been found to be highly accurate over a range of anatomical
locations.
389
CHAPTER 11
11.3.5.5. Dead time correction

Detector count rates vary between patients and can also vary within studies
performed on the same patient. In order to achieve quantitatively accurate images,
the true count rate should ideally increase linearly with increasing activity in the
FOV. Although this is usually the case at low count rates, the PET scanners
response becomes increasingly non-linear at high count rates. Individual detector
modules within a scanner require a finite period of time to process each detected
photon. If a second photon is incident upon a detector while an earlier photon
is still being processed, the secondary photon may be lost. The likelihood of
this occurring increases at high count rates and results in an effective loss of
sensitivity. This kind of count loss is referred to as dead time.
Detector dead time losses occur mostly in the detector front-end electronics.
The signal produced by an individual photon is integrated for a fixed period of
time in order to determine the position and energy of the incoming event. If
a second photon is recorded during this integration time, the two signals will
combine (pulse pile-up) in such a way as to become indistinguishable from each
other. The resulting combined signal will be rejected if it exceeds the upper level
energy discriminator. Alternatively, if it is recorded within the energy acceptance
window, it may be treated as a single event with a position somewhere between
the locations of the two individual photons. In this case, pulse pile-up contributes
to a loss of spatial resolution as well as a loss of true counts. Other sources
of dead time arise during coincidence event processing. When more than two
events occur within the coincidence time window, it is impossible to determine
the correct coincidence pair. In this circumstance, all detection events may
be discarded, contributing to dead time losses, or alternatively, all possible
coincidence events can be included, increasing the randoms component.
Dead time correction compensates for this loss of sensitivity. Corrections
are usually based upon experimental measurements of the scanners response to
a decaying source of activity. After randoms correction, residual non-linearity in
the scanners response can be attributed to dead time. An analytical model of the
scanners count rate response can be determined from these experimental data
and used for dead time correction of subsequent patient data. A global correction
factor can be applied for a particular acquisition, assuming that dead time effects
are similar for all detectors in the ring. Alternatively, different corrections can be
applied to each detector block or group of blocks. Corrections can be determined
based upon an estimate of the fraction of the acquisition period that each detector
was busy processing events and unable to process other photons. Alternatively,
the single photon rate at a particular detector can be used as input for a model
of the scanners dead time performance to estimate the magnitude of the dead
time effect. Dead time correction only compensates for count losses and does
390
not compensate for the event mis-positioning that can occur as a result of pulse
pile-up.
11.3.5.6. Image calibration
The above corrections substantially eliminate the image artefacts and
quantitative errors caused by the various physical effects that degrade PET data.
As a result, the reconstructed images reflect the activity distribution within the
FOV, within the limitations imposed by the systems limited spatial resolution.
Furthermore, these reconstructed images can be used to quantify the in vivo
activity concentration in a particular organ or tissue. Although this capability is
not always fully exploited, the potential to accurately quantify images in terms of
absolute activity concentration facilitates a range of potential applications.
After image reconstruction, including the application of the various
physical corrections, PET images have arbitrary units, typically counts per
voxel per second. Quantitative data can be extracted from the relevant parts of
the image using region of interest techniques but cannot be readily compared
with other related data such as measurements made with a radioactivity calibrator
(dose calibrator). In order to convert the PET images into units of absolute
activity concentration such as becquerels per millilitre, a calibration factor is
required. This calibration factor is experimentally determined, usually using
a uniform cylinder phantom. The cylinder is filled with a known volume of
water, to which a known amount of radioactivity is added. After ensuring the
radioactivity is uniformly distributed within the phantom, a fully corrected PET
image is acquired. The calibration factor CF can be determined using:
CF =
A p
(11.18)
V C
where
A/V is the known activity concentration (Bq/mL) within the phantom;
C is the mean voxel data (counts voxel1 s1) from a large region well
within the cylinder part of the image;
and p is the positron fraction of the radionuclide used in the calibration experiment
(typically 18F, positron fraction 0.97).
The positron fraction is a property of the radionuclide and is the fraction of
all disintegrations that give rise to the emission of a positron.
391
CHAPTER 11
The above calibration assumes that the true activity within the phantom
is accurately known. This can usually be achieved to an acceptable level of
tolerance using an activity calibrator that has been calibrated for the isotope of
interest using a long lived standard source that is traceable to a national metrology
institute. In principle, a single calibration factor can be applied to subsequent
studies performed with different isotopes as long as the positron fraction is
known. Calibrated PET images can, thus, be determined by multiplying the raw
image data by the calibration factor and dividing by the positron fraction for the
particular isotope of interest.
11.4. SPECT/CT AND PET/CT SYSTEMS
11.4.1. CT uses in emission tomography
SPECT and PET typically provide very little anatomical information,
making it difficult to precisely localize regions of abnormal tracer accumulation,
particularly in oncology studies where disease can be widely disseminated.
Indeed, it is often the case that the more specific the radiopharmaceutical, the
less anatomical information is available to aid orientation. Relating radionuclide
uptake to high resolution anatomic imaging (CT or MRI (magnetic resonance
imaging)) greatly aids localization and characterization of disease but ideally
requires the two images to be spatially registered. Retrospective software
registration of images acquired separately on different scanner systems has
proved to be effective in certain applications, notably for brain studies where
rigid body assumptions are realistic. However, for most other applications,
the rigid body assumption breaks down and the registration problem becomes
much more difficult. Combined scanner systems, such as SPECT/CT and PET/
CT, provide an alternative solution. The advantage of this hardware approach is
that images from the two modalities are inherently registered with no need for
further manipulation. Of course, this assumption can become unreliable if the
patient moves during data acquisition, but, in general, combined scanner systems
provide an accurate and convenient method for achieving image registration.
In addition to the substantial clinical benefit of registered anatomical and
functional images, the coupling of CT with SPECT and PET systems provides
an additional technical benefit. Although radionuclide sources have been used
for attenuation correction, the availability of co-registered CT is particularly
advantageous for this purpose. In the case of SPECT, the main advantages of
CT based attenuation correction are greater accuracy and reliability compared to
radionuclide sources, while in PET, the main advantage is an effective reduction
in the overall duration of the scanning procedure owing to the speed with which
392
CT images can be acquired. In addition, there are a number of other subsidiary

benefits to the introduction of CT to SPECT and PET systems. Radionuclide
transmission sources and their associated motion mechanisms are somewhat
cumbersome, particularly in the case of SPECT, and the addition of the CT allows
this component to be removed from the design. Elimination of the transmission
source from PET systems enabled the patient port size to be enlarged, allowing
larger patients to be accommodated. Other benefits include use of the CT
subsystems localizing projection image to aid patient positioning, particularly
for single bed-position PET scans. In addition, the potential of acquiring both a
radionuclide study and diagnostic quality CT in the same scanning session has
advantages in terms of convenience. In terms of quantitative image analysis,
the availability of registered CT along with a radionuclide study can sometimes
be useful for region of interest definition, particularly in research applications.
Although giving rise to a number of significant benefits, it should be noted that
the addition of CT to SPECT and PET instrumentation has led to an appreciable
increase in the radiation dose received by patients undergoing radionuclide
imaging procedures.
11.4.2. SPECT/CT
In many respects, SPECT might be expected to benefit more than PET
from the addition of registered CT. SPECT has lower spatial resolution than
PET. Many SPECT tracers are quite specific and often do not offer the kind of
anatomical orientation that is provided by normal organ uptake with PET tracers
such as fluorodeoxyglucose (FDG). Radionuclide transmission scanning is more
awkward in SPECT compared to PET because gamma cameras need to be
capable of multiple flexible modes of acquisition. Despite these considerations,
the adoption of combined SPECT/CT instrumentation (Fig.11.44) has been
slower than that of PET/CT. Cost considerations no doubt contribute and there
remains uncertainty about the level of CT performance that is required for a
SPECT/CT system.
FIG.11.44. Clinical SPECT/CT systems.
393
CHAPTER 11
Early SPECT/CT designs, including successful commercial offerings,

coupled SPECT with low performance CT. The SPECT subsystem was placed
in front of the CT and data were acquired sequentially with the patient being
translated between gantries by a common patient bed. Sequential as opposed to
simultaneous data acquisition was required due to cross-talk between the CT
and SPECT subsystems and was quickly established as the standard mode of
operation for both SPECT/CT and PET/CT. Low power X ray tubes and slow
rotation speeds meant that the CT component was by no means optimized for
diagnostic quality imaging. The aim was to provide attenuation correction and a
much needed anatomical context for the SPECT, while maintaining a relatively
low cost.
The desire for improved CT image quality has led to the introduction of
SPECT/CT systems that incorporate a high performance CT component with
capabilities comparable to dedicated CT scanners. With this development,
SPECT/CT now benefits from substantially improved CT image quality, faster
data acquisition and a broader range of CT protocols. A number of different
multi-detector CT slice configurations are available, as well as alternative designs
including those based upon flat panel detectors and improvements of the original
low cost, non-diagnostic CT. This broad range of CT capability may reflect a
diversity of opinion about the role of combined SPECT/CT in clinical practice.
11.4.3. PET/CT
Concurrent technical developments in both PET and CT were exploited in
the development of the combined PET/CT system [11.4]. In PET, new detector
materials and approaches to image reconstruction increased spatial resolution
and improved image statistical quality. In CT, the introduction of spiral scanning
and multi-detector technology enabled extremely fast data acquisition over an
extended FOV. These developments meant that the combined PET/CT system
was well positioned to take advantage of the growing evidence that PET imaging
with FDG could play a valuable role in oncology (Fig.11.45). The addition
of spatially registered anatomical information from the CT component of the
combined scanner provided the impetus for widespread acceptance of PET in
oncology and has driven the rapid growth of PET/CT instrumentation. PET/CT
has now been rapidly accepted by the medical community, so much so that
stand-alone PET systems are no longer being developed by the major commercial
vendors.
Early PET/CT systems were not closely integrated and consisted of
separately developed PET and CT components that operated independently of
each other. Given that PET/CT acquisitions occur sequentially, as opposed to
simultaneously, this approach was reasonable but it meant that multiple computer
394
systems were required and the user interface could be awkward. In addition, the
availability of the CT for attenuation correction meant that the PET transmission
scanning system was somewhat redundant. Subsequent designs removed the PET
transmission sources and moved the two subsystems towards greater integration.
In some cases, this meant a more compact system with a continuous patient
tunnel. In other cases, the PET and CT gantries were separated by a gap which
allowed greater access to the patient during scanning. Removing the transmission
scanning system also provided scope for increasing the size of the patient port,
so as to accommodate larger patients. This was further achieved by removing
the septa from the PET subsystem and decreasing the size of the end-shields
used to reject out of FOV radiation. Although the PET and CT detectors remain
separate subsystems, many software functions of a modern PET/CT system run
on a common platform, including a common patient database containing both
PET and CT data.
FIG.11.45. Coronal images from a whole body fluorodeoxyglucose PET/CT study. (a) The
PET data are shown in inverse grey scale. (b) The same PET data are shown in a colour scale,
superimposed on the CT in grey scale.
The CT component of combined PET/CT systems continues to evolve as

greater multi-detector capability becomes available. The initial PET/CT design
395
CHAPTER 11
involved single slice CT, whereas later systems have incorporated 4, 16, 64 or
greater slice CT. Advanced CT detector capability allows for extended volume
coverage in a single breath-hold or contrast phase and also facilitates a range
of rapid diagnostic CT protocols, particularly those intended for cardiology. The
level of CT capability required by a combined PET/CT system depends on the
extent to which the system will be used for diagnostic quality CT. For many PET/
CT oncology studies, state of the art CT is not necessary and low dose protocols
are favoured. In these cases, the X ray tube current is reduced and intravenous
contrast would typically not be administered. In addition, for whole body studies,
data would not be acquired under breath-hold conditions in order to improve
spatial alignment of the CT with the PET data that are acquired over multiple
respiratory phases.
As far as CT based attenuation correction is concerned, the main advantage
for PET is not so much improved accuracy, as 68Ge transmission sources are
perfectly adequate when obtained with sufficient statistical quality. The main
advantage of CT based attenuation correction is the speed with which the data
can be acquired. This is particularly important for whole body PET studies
where transmission data are needed at each bed position, requiring extended scan
durations to achieve adequate statistical quality. With multi-detector CT scanners,
low noise images can be acquired over the whole body in only a few seconds.
Replacing radionuclide transmission scans with CT had the effect of substantially
reducing the overall duration of scanning procedures, particularly those requiring
extended axial coverage. In turn, shorter scans increase patient comfort and
reduce the likelihood of motion problems. With the shorter scan times, arms
up acquisition is better tolerated by patients, leading to reduced attenuation
and improved image statistical quality. In addition to these methodological
considerations, shorter scan durations allow for more patient studies in a given
time period and, thus, more efficient utilization of the equipment.
Despite the relatively rapid scanning afforded by modern PET/CT systems,
the PET component still typically requires manyminutes of data acquisition and
management of patient motion continues to be a problem. Motion can potentially
cause misalignment of the PET and CT images, degradation of image spatial
resolution and the introduction of characteristic artefacts. Although the CT
images are acquired independently of the PET, the PET images involve CT based
attenuation correction and are, thus, particularly susceptible to patient motion
between the two scans. The types of motion usually encountered include gross
movement, often of the arms or head, and respiratory motion. The former is hard
to correct retrospectively, although external monitoring devices such as camera
based systems can potentially provide solutions for head motion. External
monitoring devices can also help reduce problems due to respiratory motion.
Respiratory gated PET can reduce motion blurring by reconstructing images
396
from data acquired only during particular phases of the respiratory cycle. This
can be further refined by including a motion model into the PET reconstruction
and incorporating CT data also acquired under respiratory control.
REFERENCES
[11.1] ANGER, H.O., Scintillation camera, Rev. Sci. Instrum. 29 (1958) 2733.
[11.2] CHERRY, S.R., SORENSON, J.A., PHELPS, M.E., Physics in Nuclear Medicine,
[11.3] BAILEY, D.L., TOWNSEND, D.W., VALK, P.E., MAISEY, M.N., Positron Emission
Tomography: Basic Sciences, Springer, London (2005).
[11.4] BEYER, T., et al., A combined PET/CT scanner for clinical oncology, J. Nucl. Med.
41 (2000) 13691379.
397
CHAPTER 12
COMPUTERS IN NUCLEAR MEDICINE
J.A. PARKER
Division of Nuclear Medicine and Department of Radiology,
Beth Israel Deaconess Medical Center,
Harvard Medical School,
Boston, Massachusetts,
12.1. PHENOMENAL INCREASE IN COMPUTING CAPABILITIES
I think there is a world market for about five computers this remark is
attributed to Thomas J. Watson (Chairman of the Board of International Business
Machines), 1943.
12.1.1. Moores law
In 1965, Gordon Moore, a co-founder of Intel, said that new memory chips
have twice the capacity of prior chips, and that new chips are released every
18 to 24 months. This statement has become known as Moores law. Moores law
means that memory size increases exponentially. More generally, the exponential
growth of computers has applied not only to memory size, but also to many
computer capabilities, and since 1965, Moores law has remained remarkably
accurate. Further, this remarkable growth in capabilities has occurred with a
steady decrease in price.
Anyone who has even a little appreciation of exponential growth realizes
that exponential growth cannot continue indefinitely. However, the history of
computers is littered with experts who have prematurely declared the end of
Moores law. The quotation at the beginning of this section indicates that future
growth of computers has often been underestimated.
12.1.2. Hardware versus peopleware
The exponential growth of computer capabilities has a very important
implication for the management of a nuclear medicine department. The growth
in productivity of the staff of a department is slow, especially when compared to
the growth in capabilities of a computer. This means that whatever decision was
398
made in the past about the balance between staff and computers is now out of
date. A good heuristic is: always apply more computer capacity and less people
to a new task. Or stated more simply, hardware is cheap, at least with respect
to what you learned in training or what you decided last time you considered the
balance between hardware and peopleware.
12.1.3. Future trends
In the near future, the increase in personal computer capability is likely
to be due to an increase in the number of central processing units on a single
processor chip (cores) and in the number of processing chips in a single computer.
Multiple processing units have been a key feature of supercomputers for many
years. Coordinating the large number of processors is often a bottleneck in the
application of supercomputers for general purpose computing. Supercomputers
have generally been applied to specific tasks, not to general purpose computing.
The trend towards more cores in personal computers has also suffered from
this bottleneck. Existing applications often run marginally faster on multicore
computers than they do on a single core.
Multi-threaded programming, which has recently received more
attention, ameliorates this bottleneck. Multi-threading is an efficient method of
synchronizing subtasks that can be computed independently in parallel. Image
processing, where different parts of the image can be processed independently, is
well suited to multi-threading. Reworking just the intensive processing portions
of the software as a multi-threaded application can greatly improve the overall
speed on a multicore machine. In fact, several basic image processing packages
are already multi-threaded. Thus, with relatively limited updating, nuclear
medicine software should be able to take advantage of the trend towards multiple
processors in a single computer.
An area where multiple processing units are currently used in personal
computers is in graphical processing units (GPUs). GPUs, which perform the
same operation on multiple parts of an image simultaneously, are classified as
single instruction, multiple data processors. These units have traditionally been
thought to be very difficult to program, but recently some programming tools are
making them somewhat more readily accessible. They are still very difficult to
program in comparison to multi-threading; however, the improved programming
tools should make these processors more common for the most computing
intensive data processing tasks. They are more likely to be used in the front-end
computers within the imaging devices (seeChapter11) than in workstations and
servers that will be the focus of this chapter.
399
CHAPTER 12
12.2. STORING IMAGES ON A COMPUTER

12.2.1. Number systems
12.2.1.1. Decimal, binary, hexadecimal and base 256
Everyone is familiar with the Arabic or decimal number system, which
uses ten symbols, 09. Computer circuits are best at dealing with two values,
fully on and fully off. The binary number system has two values, 0 and 1. These
values can represent on/off, high/low or true/false. From right to left, the digits
in a binary number represent ones, twos, fours, eights, sixteens, etc. Computers
have become much better at performing conversions to decimal numbers, so that
users generally do not have to worry about binary numbers; however, images
are an exception. It is still useful to know something about binary numbers to
understand image representation at a fundamental level.
Binary numbers are natural for computers, but they are very unnatural for
humans. A compromise between humans and computers is the hexadecimal or
base-16 number system. The decimal number, 1000, is represented by 3E8 in
hexadecimal, which is much easier to remember than 1111101000 in binary.
The hexadecimal number system uses 16 symbols: 09 are used for the first ten
symbols and AF are used for the last six symbols. From right to left, the digits in
a hexadecimal number represent 1s, 16s, 256s, etc.
The value of a decimal digit is 10n, where n is the position. The value of a
binary digit is 2n, and the value of a hexadecimal digit is 16n. Four binary digits
represent the same value as one hexadecimal digit: 24=16. For this reason,
one can easily convert back and forth between binary and hexadecimal, but not
between binary and decimal.
Another interesting number system is used with the Internet (Table12.1).
It uses a base 256 number system, but rather than inventing a whole new set
of symbols, it uses the corresponding decimal numbers as symbols. Since the
decimal numbers from 0 to 255 vary in length, it needs to use periods to separate
the digits. From right to left, the digits in an Internet address represent 1s, 256s,
2562s, etc. The Internet number system is the most convenient compromise
between the computers preference for binary numbers and peoples familiarity
with the decimal system. An internet address can be considered to be a base
256 (28) number. The digits are separated by periods. Digit values are given by
their decimal equivalents. One internet digit equals two hexadecimal digits or
eight binary digits (see Table 12.1).
400
12.2.1.2. Kilo, mega, giga, tera

In the decimal system, kilo, mega, giga and tera are used to represent 1000,
1000000, 1000000000 and 1000000000000, respectively. Using scientific
notation, these numbers are 103, 106, 109 and 1012, respectively. Twelve binary
digits can be used to represent roughly the same range of numbers as three
decimal digits: 210=1024 and 103=1000. Therefore, these terms have been
appropriated by the computer community to mean 210, 220, 230 and 240.
TABLE12.1. Internet address number system
32 binary digits
Binary
Hexadecimal
0001
1000
0000
1001
1111
0011
Internet address
24.
9.
243.
0100 1110
4
E
78.
12.2.2. Data representation

Digital images are composed of individual picture elements, pixels, which
represent a single point in the image. Each pixel is represented by a number or a
series of numbers. There are several methods of representing each number.
12.2.2.1. Integer numbers
A group of binary digits (bits) can be interpreted in several different ways.
The simplest is as a positive integer, numbers 0, 1, 2, 3, etc. The number of
bits determines how large a number can be stored. Four bits, one hexadecimal
digit, can represent 015. Eight bits can represent two hexadecimal digits,
values 0255. Eight bits are called a byte.
A byte is usually the smallest unit of storage that is used for a pixel. It is
fairly limited, both in terms of the number of counts that can be collected in one
pixel and in terms of how many colours can be specified by 1 byte. Given the
architecture of modern computers, it makes more sense to use a number of bytes,
e.g. 2 bytes (16 bits), 3 bytes (24 bits) and 4 bytes (32 bits). Table12.2 shows the
number of counts that can be represented using each of these formats. In nuclear
medicine, a 2 byte image is sometimes called a word image. More generally,
word does not have a fixed meaning; it varies with the computer architecture.
It is a less ambiguous way to describe larger integer values as 2 byte, 4 byte, etc.
401
CHAPTER 12
A further complication is that pixels are sometimes represented by

non-integral numbers of bytes. For example, computed tomography (CT) pixels
are often 12 bits. Within computers, these pixels are often stored with 2 bytes,
where the unused bits are set equal to zero. Some image formats even use a
non-integral number of bytes per pixel.
TABLE12.2. THE NUMBER OF COUNTS THAT CAN BE REPRESENTED
IN DIFFERENT IMAGE FORMATS
Bits
Range
Number of values
1 bit
01
1 byte
0255
256
2 bytes
16
065535
64 k
3 bytes
24
016777215
16 M
4 bytes
32
04294967295
4G
5 bytes
40
01099511627 776
1T
where
Symbol
Prefix
Power of 2
Value
Power of 10
kilo
210
1024
~103
mega
220
1048567
~106
giga
230
1073741824
~109
tera
240
1099511627776
~1012
12.2.2.2. Signed versus unsigned numbers

Positive integers are called unsigned integers. In 2s compliment
arithmetic, negative integers, 1, 2, 3, etc., are represented by the largest, next
largest, etc. binary numbers. The sign of the number is determined from the value
of the highest order bit. If a signed number is displayed by mistake as an unsigned
number, it is shown as if it had very high activity. Understanding the difference
between signed and unsigned numbers makes it relatively easy to diagnose and
know-how to solve this problem.
402
12.2.2.3. Floating point and complex numbers

Floating point representation is analogous to scientific notation where a
number is represented by a mantissa times 10n, where n is called the exponent.
Instead of 10n, computer representation uses 2n. Using the Institute of Electrical
and Electronics Engineers (IEEE) standard, both the number and the exponent are
signed integers. For the single-precision, 32-bit IEEE standard, the magnitude of
the largest number is about 3.41038 and the magnitude of the smallest non-zero
number is about 1.21038. For the double-precision, 64-bit IEEE standard, the
magnitude of the largest number is about 1.210308 and the magnitude of the
smallest non-zero number is about 2.210308.
Floating point numbers are not generally used for raw nuclear medicine
data; however, during processing, they can be quite useful both to represent
fractions and because some processing involves large intermediary values.
Complex numbers consist of two parts, a real part and an imaginary part.
They are written x + iy where x is the real part and iy is the imaginary part. The
symbol i stands for 1. In mathematics, the symbol j is often used instead of i.
Each complex number is represented as two floating point numbers. Complex
numbers only come up in nuclear medicine during processing. However, in
magnetic resonance imaging (MRI), complex numbers are the most logical
representation for the raw signals that come from the magnet.
12.2.2.4. Byte order
Memory is addressed by byte. When a number is stored using 4 bytes, then
successive numbers start at addresses 0, 4, 8, etc. Confusion can arise because
computer manufacturers did not adopt a standard way of assembling the bytes
into a number. From least significant to most significant, some manufactures
assembled 4-byte numbers using address 0 followed by address 1, e.g. 0123;
other manufacturers assembled 4-byte numbers in the order 3210. The former is
sometime called big-endian, the big address is at the end the big values come
first; the latter is sometimes called little-endian.
12.2.3. Images and volumes
Images can be represented by 2D functions. If x and y are taken to be
horizontal and vertical, a 2D function, f(x, y), gives the value of the image at
each point in the image. Ever increasingly, imaging equipment produces not a
single image, but a 3D volume of data, e.g. right to left, anterior to posterior
and caudal to cephalic. A volume of data can be represented as a 3D function,
f(x, y, z).
403
CHAPTER 12
2D, 3D, 4D, etc. are often used loosely in medicine, but it can be
insightful to clearly understand the dimensions involved in an application.
A relevant example is human vision. A single human eye can see in only two
dimensions; the retina is a 2D structure. From parallax as well as other
physiological inputs, it is possible for people with binocular vision to perceive
the depth of each point in the visual image. The most appropriate model of the
perceived image is a multi-valued function with value intensity, hue, saturation
and depth. In this model, the function is 2D. This 2D function represents a
surface in a 3D volume. Over time, themind is able to construct a 3D model
from sequential 2D surfaces.
On a more basic level, optics models an electromagnetic signal going
through an aperture as a complex 2D function. In addition to amplitude
information, which can be sensed by the eye, the function also has phase
information. Holography takes advantage of phase information, allowing the
observer to see around objects if there is no other object in the way. However,
basic physics limits the amount of information to a sparse (essentially 2D) set of
data contained within the 3D space.
One of the challenges of data visualization is to facilitate the input of
3D data using the 2D channel afforded by the visual system. Often, one
dimension is mapped into time using cine or a mouse to sequence through a stack
of images. Rendering such as a reprojection or a maximum intensity projection
can also help by providing an overview or by increasing the conspicuousness of
the most important features in the data. Both of these visualization methods also
use a sequence of images to overcome the limitations of the 2D visual channel.
12.2.3.1. Continuous, discrete and digital functions
The real world is usually modelled as continuous in space and time.
Continuous means that space or time can be divided into infinitesimally small
increments. A function f(x) is said to be a continuous function if both the
independent variable x and the dependent variable f are represented by continuous
values. The most natural model for the distribution of a radiopharmaceutical in
the body is a continuous 3D function.
An image that is divided into pixels is an example of a discrete function.
The independent variables x and y, which can only take on the values at particular
pixel locations, are digital. The dependent variable, intensity, is continuous.
A function with independent variable(s) that are digital and dependent variable(s)
that are continuous is called a discrete function.
A computer can represent only digital functions, where both the
independent and dependent variable(s) are digital. Digital values provide a good
model of continuous values if the coarseness of the digital representation is small
404
with respect to the standard deviations of the continuous values. For this reason,
digital images often provide verisimilar representations of the continuous world.
Nuclear medicine is intrinsically digital in the sense that nuclear medicine
imaging equipment processes scintillation events individually. The original Anger
scintillation cameras produced analogue horizontal and vertical position signals,
but modern scintillation cameras process the signals digitally and the output is
digital position signals. Most positron emission tomography (PET) cameras have
discrete crystals, so that the lines of response are intrinsically digital.
12.2.3.2. Matrix representation
The surface of the gamma camera can be visualized as being divided into
tiny squares. An element in a 2D matrix can represent each of these squares.
A 3D matrix can represent a dynamic series of images or single photon emission
computed tomography (SPECT) data collection. A 3D matrix is equivalent
to a 3D digital function, f[x, y, z]. Both representations are equivalent to the
computer program language representation, f[z][y][x].
The lines of response in a PET camera without axial collimation are more
complicated. Perhaps the easiest way to understand the data is to consider a ring
of discrete detectors. The ring can be represented as a 2D array one axial
and one tangential dimension. Any line of response (LOR) can be defined by
two crystals, and each of those two crystals is defined by two dimensions. Thus,
each LOR is defined by four dimensions. A 4D matrix can represent the lines of
response (not all crystal pairs are in coincidence, so the matrix is sparse, but that
is a detail).
It is particularly simple for a computer to represent a matrix when the
number of elements in each dimension is a power of two. In this case, the x, y and
z values are aligned with the bits in the memory address. Early computer image
dimensions were usually powers of two. Modern programming practice has
considerably lessened the benefit of this simple addressing scheme. However,
where hardware implementation is a large part of the task, there is still a strong
tendency to use values that are a power of two.
12.3. IMAGE PROCESSING
This section will present an introduction to the general principles of image
processing.
405
CHAPTER 12
12.3.1. Spatial frequencies

12.3.1.1. Vision and hearing
Humans conceptualize images in terms of what they perceive. Thinking
of images in terms of spatial frequencies is not natural. However, the relations
between perception, f(x, y), and a spatial frequency representation, F(kx, ky), can
be understood by appealing to an analogy to hearing. Sound waves striking the
ear-drum can be represented by a function of time, f(t). A pure tone is a sinusoidal
variation in pressure as a function of time. Humans do not think of a sinusoid
when they hear a pure tone; it is much more natural to think of a pure tone in
terms of a musical note.
The cochlea transforms sinusoidal changes in pressure over time into firing
of the single nerve in the auditory nerve that represents this tone. The pressure
signal, f(t), is transformed to frequencies. If represents the frequencies, the
transformed signal can be represented by the function F(). It is recalled that
angular frequency is 2 times frequency. In this simple case, F() is an impulse
at the frequency corresponding to the tone. The process that the cochlea performs
can be represented mathematically by a Fourier transform. The time domain
representation, f(t), of the pressure signal is transformed into a frequency domain
signal, F(), carried on the auditory nerve.
Humans hear tones; they do not feel the pressure waves. Thus, the frequency
domain is more natural. Since sound waves are part of every ones early education,
they are less natural than tones, but do not seem foreign. Spatial frequencies do
seem foreign to most people. However, the relationship between f(t) and F() is
exactly analogous to the relationship between f(x, y) and F(kx, ky). This analogy
may help physicians and other non-mathematically oriented health professionals
conceptualize the relationship between space and spatial frequencies.
12.3.1.2. Spatial sinusoids
A pure spatial frequency corresponds to a sinusoidal function in the
spatial domain. Sinusoidal means a sine, a cosine or a combination of the two.
Figure12.1 shows a number of different sinusoidal functions.
The spatial frequency variables kx and ky are often given in terms of the
cycles per pixel. A signal that has successive maximums andminimums in
adjacent pixels will have one complete cycle in two pixels or 0.5 cycles per
pixel. The spatial frequency scale is often shown from 0 to 0.5 cycles per pixel,
where 0 cycles per pixel represents a constant value in the image domain and 0.5
cycles per pixel represents an image which varies from maximum tominimum in
2 pixels.
406
FIG.12.1. The upper left quadrant is a sinusoid with a single frequency in the x direction;
the upper right quadrant is a sinusoid with twice the frequency; the lower left quadrant is a
sinusoid with a single frequency in the y direction; and the lower right quadrant is a single
sinusoid which varies in both the x and y direction.
12.3.1.3. Eigenfunctions
Eigenfunctions or eigenvectors can be a difficult mathematical concept.
They are often presented as an advanced concept. However, eigenfunctions are
actually a basic concept. They are relatively easy to understand if the detailed
mathematics is avoided.
Eigenfunctions are the natural functions of the system. For example,
an eigenfunction of a swing is a sinusoidal back and forth motion. The swing
407
CHAPTER 12
naturally wants to go back and forth in a sinusoidal motion. The output of a system
when the input is an eigenfunction is a scaled version of the input. The system
does not change the shape of the eigenfunction, it just scales the magnitude. For
example, if a pure tone is put into a good audio amplifier, the same tone comes
out, only amplified.
The model implied by these descriptions is that there is a system that has
an input and an output. In the first case, the system is a swing; in the second, the
system is an audio amplifier. It should be noted that eigenfunctions are properties
of systems. A more relevant example is where the system is a gamma camera, the
input is the distribution of radioactivity from a patient and the output is an image.
The swing and the amplifier have sinusoids as eigenfunctions. In fact, a
large class of systems have sinusoids as their eigenfunctions. All linear-timeinvariant or linear-shift-invariant systems have sinusoids as their eigenfunctions.
Time-invariant or shift-invariant mean that the properties of the systems do not
change with time or with position. Real imaging systems are rarely linear-shiftinvariant systems, but there is often a region over which they can be modelled as
a linear-shift-invariant system, so the sinusoids will be very useful.
A common use of eigenfunctions in nuclear medicine is in measuring
the modulation transfer function (MTF). The MTF is measured by imaging a
spatial sinusoid and seeing how well the peaks and valleys of the sinusoid (the
modulation) are preserved (transferred by the system). If they are completely
preserved, the modulation transfer is 1. If the peaks and valleys are only half
of the original, the modulation transfer is 0.5. The typical bar phantom is an
approximation of a sinusoidal function.
The key point for this section is that a spatial sinusoid is an eigenfunction
of the imaging system, at least over some region. The image has the same shape
as the input, only the amplitude (modulation) of the sinusoid is altered. The basic
properties of the system can be determined by measuring the modulation transfer
of the eigenfunctions. Linearity means that the effect of the imaging system on
any signal that is a combination of eigenfunctions can then be determined as a
scaled sum of eigenfunctions.
12.3.1.4. Basis functions
A function, f(t), can be made up of a sum of a number of other functions,
gk(t). For example, a function can be made from a constant, a line passing through
the origin, a parabola centred at the origin, etc. This function can be written as the
sum of K powers of t:
f (t ) =
408
K 1
k=0
a k t k (12.1)
These functions are just polynomials. The terms tk in the polynomial are
basis functions, with coefficients ak. Selecting different coefficients ak, a large
number of functions can be represented.
Usually, the powers of the independent variable tk seem like they are the
most important part of a polynomial, and in many ways they are. However, it will
be useful to shift the focus from the powers of t to the coefficients ak. To make
a new polynomial, the key is to select the coefficients. From this viewpoint, the
tk terms are just placeholders.
Extending this point of view, the coefficients can be thought of as a discrete
function F[k]. The polynomial function could be rewritten:
f (t ) =
K 1
k=0
F [k ]t k (12.2)
This polynomial function provides a method of transforming from the

function of coefficients to the function of time. This process is called evaluation
of the polynomial. The two functions, the function of coefficients and the
function of time, are equivalent representations in the sense that it is possible
to transform the function of the coefficients into the function of time using this
transformation. It is also possible to select K points in f(t) and transform them to
F[k]. This process, called interpolation, is beyond the scope of this chapter.
For this discussion, the key point is that the coefficient and the time
representations can be converted back and forth with the processes of evaluation
and interpolation. f(t) can be described as a function in the time domain. F[k]
can be described as a discrete function in the coefficient domain. These two very
different functions represent the same thing in the sense that using evaluation or
interpolation, it is possible to convert back and forth between the representations.
12.3.1.5. Fourier domain k-space
When learning about a new way of representation of a familiar concept, it
is natural to think of the familiar concept as the real representation and the new
representation as artificial. The time or space domains are the real domains,
the frequency or the spatial-frequency domains are the artificial or transform
domains. However, they are really equivalent representations. One way to
understand this is to think about hearing and sight. The natural domain for sight
is the space domain, but the natural domain for hearing is the frequency domain.
In one case, the natural domain is frequencies and the other the natural domain is
just the opposite. The equivalence of the two domains is called duality. Duality is
a difficult concept. Electrical engineers, who work with Fourier transforms all the
409
CHAPTER 12
time, still think of the time domain as the real domain and the frequency domain
as the transform domain.
One of the best examples of duality is MRI. Chemists used to working with
nuclear magnetic resonance think of the time signal as the natural signal and
the frequency signal as the transform signal. Imagers think of the image from a
magnetic resonance imager as the natural signal and the spatial frequencies as the
transform domain. However, due to the gradients, the frequency signal gives the
spatial representation and the time signal is the spatial-frequency signal. There is
no real domain and transform domain; it all depends on the point of view.
12.3.1.6. Fourier transform
The Fourier transform equations can be written compactly using complex
exponentials:
F ( ) =
f (t ) =
f (t)e
1
2
i t
dt (12.3)
F ( )e
i t
d (12.4)
where i is 1 .
The first equation (Eq.(12.3)) from the time or space domain to the
frequency domain is called Fourier analysis; the second equation (Eq.(12.4)),
going from the frequency domain to the time or space domain, is called Fourier
synthesis. Fourier analysis is analogous to polynomial interpolation; Fourier
synthesis is analogous to polynomial evaluation. The relation of these equations
to the sine and cosine transforms can be seen by substituting for the complex
exponential using Eulers formula:
e i t = cos( t ) + i sin( t ) (12.5)
These equations have been written in terms of time and frequency.

Analogous equations can be written in terms of space and spatial frequency. For
the 2D case:
F (k x , k y ) =
410
f (x, y)e
i( k x x+k y y)
dx dy (12.6)
f ( x, y) =
1
2
F (k , k )e
x
i( k x x+k y y)
dk x dk y (12.7)
The common use of the letter k with a subscript for the spatial frequency
variable has led to the habit of calling the spatial frequency domain the k-space,
especially in MRI. These equations are exactly analogous to the time and
frequency equations with t replaced by x, y, and replaced by kx, ky. In the case
of three dimensions, t is replaced by x, y, z, and by kx, ky, kz.
The previous equations refer to continuous functions. The limits of
integration of the integrals were not specified, but in fact, the limits are assumed
to be to +. However, computer representation is digital. Computer
representation is often thought of as discrete, not digital, since the accuracy of
representation of numbers is often high, so that the quantification effects can be
ignored (seeSection 12.2.3.1). The Fourier transform equations in discrete form
can be written as:
F [k ] =
f [n] =
1
N
N 1
n=0
f [n]e i2 kn / N (12.8)
N 1
k=0
F [k ]e i2 kn / N (12.9)
In image processing, the unit of n is often pixels, and the frequency unit k is
often given as a fraction, cycles/pixel. The space variable n runs from 0 to N 1;
the spatial frequency variable k runs from 0.5 cycles/pixel to (but not including)
+0.5 cycles/pixel in steps of 1/N.
12.3.1.7. Fourier transform as a correlation
Some understanding of how the Fourier transform pair works can be
obtained by noting the analogy between these equations and a correlation. The
correlation coefficient is written:
r=
x y (12.10)
( x y )
i i
2
i
2
i
where x and y are the two variables, and i indices over the number of samples.
The denominator is just normalization, so that r ranges from 1 to 1. The action
is in the numerator. It should be noted that the key feature of a correlation is that
it is the sum of the products.
411
CHAPTER 12
There is an analogy between the correlation equation and the Fourier

transform equation. The index i is analogous to the variable t; xi is analogous to
f(t); yi is analogous to the sinusoid. The Fourier transform equation determines
how much a function is like the sinusoid, just as the correlation coefficient
determines how much two variables are alike.
12.3.2. Sampling requirements
To represent a function, f[n], with N points, N frequency values, F[k], are
needed; to represent a frequency domain signal, F[k], with N points, N time or
space values, f[n], are needed. As a general heuristic, if a signal has N points in
one representation, N points are needed in another representation. This heuristic
assumes that the N points are independent.
A typical property of real systems is that they can only handle signal
variations up to a particular maximum frequency. Thus, real signals often have
a limited frequency content. The sampling theorem says that if a signal only
contains frequencies up to a maximum of f, then it can be exactly reproduced
from samples that are taken every 2/f per second. Sampling needs to be at
twice the highest frequency. If the sampling is not fast enough, then the high
frequencies appear to occur at lower frequencies. This process that occurs by
sampling a frequency that is too low is called aliasing.
12.3.3. Convolution
Convolution can be used to describe any linear-time-invariant or linearshift-invariant system. The input, the system and the output are each described
by a function. The system function h(t) is defined to be the output g(t) when the
input f(t) is equal to a delta function (t).
In general, the input f(t) can be conceptualized as a sequence of delta
functions, (t ), scaled by the input at time f(). Each delta function produces
a component of the output that is the system function, h(t ), which has been
shifted by a time equal to the time of the input and scaled by the magnitude of
f(). As the system is time-invariant, the same response h(t) can be used for an
input at any time. As the system is linear, the inputs at different times can be
considered separately and the separate outputs can be added. The formula for
convolution is:
g (t ) =
412
h(t ) f ( ) d (12.11)
12.3.3.1. Shift, scale, add

Convolution can be summarized as shift, scale, add shift the system
function by an amount equal to the time of the input, scale by the size of the input
and add all of the resulting shifted and scaled system functions. Convolution
can be used to describe a timeactivity curve from a region of interest. The
arterial concentration is the input, the system function is the timeactivity curve
after arterial injection of a bolus of activity, and the timeactivity curve after
intravenous administration is the convolution of the arterial concentration and the
system function.
Convolution is not limited to 1D examples. The system function for an
Anger camera is the image obtained in response to a point source of activity.
A point source can be modelled as a 2D delta function. The image of a point
source will be a blob where the size of the blob is related to the cameras
resolution. The blob can be shifted to every point in the field of view (FOV); it
can be scaled for the amount of activity at each location; and all of the blobs can
be added up. This process shift, scale, add will produce the final output
image.
The formula for 2D convolution is:
g( x, y) =
h(x x', y y', z z') f (x', y') dx' dy' (12.12)
In three dimensions:
g( x, y, z) =
h(x x', y y', z z') f (x', y', z') dx' dy' dz' (12.13)
12.3.3.2. Mapping convolution to multiplication

In Section 12.3.1.4, on basis functions, it was noted that interpolation
could be used to transform from a set of values of a signal to a representation
in terms of the coefficients of a polynomial function, and that evaluation could
be used to transform from the coefficients of a polynomial function into values.
The coefficients can be thought of as a function in the coefficient domain and
the values can be thought of as a function in the value domain. The process of
multiplying two polynomials involves keeping track of coefficients of the t0s, t1s,
t2s, etc. The ts are essentially placeholders.
In polynomial multiplication, the coefficients of the first polynomial are
shifted by the exponent of the second polynomial, scaled by the coefficients of
the second polynomial, and the products are added. The process shift, scale,
413
CHAPTER 12
add is convolution. Multiplying polynomials is complicated. However,

the process is much simpler in the value domain. Two signals are multiplied
in the value domain by simply multiplying their values. In the value domain,
multiplication of two signals is implemented by multiplying their values; in the
coefficient domain, multiplication of two signals is implemented by convolution.
Transformation of the complicated process of convolution in one domain into the
simple process of multiplication in the other domain is one of the key properties
of representing signals in terms of a set of basis functions.
In an exactly analogous fashion, the Fourier transform maps convolution
into multiplication. The Fourier transform of the convolution of two functions
is equal to the product of the Fourier transforms of the functions. Symbolically:
h(t ) f ( ) d F ( )H ( ) (12.14)
where the symbol represents Fourier transformation, the left side shows the
time domain operations, and the right side shows the Fourier domain operation.
The complicated integral in the time domain is transformed into a simple
multiplication in the frequency domain.
At first, it may seem that the Fourier transform operation is just as
complicated as convolution. However, since the fast Fourier transform algorithm
provides an efficient method for calculating the Fourier transform, it turns out that,
in general, the most efficient method of performing convolution is to transform
the two functions, multiply their transforms and do an inverse transform of the
product. Calculation of convolutions is one of the major practical applications of
the Fourier transform.
12.3.4. Filtering
The word filtering refers to processing data in the Fourier domain by
multiplying the data by a function, the filter. Conceptually, the idea is that the
frequency components of a signal are altered. The data are thought of in terms
of their frequency or spatial frequency content, not in terms of their time or
space content. The most efficient process in general is (i) Fourier transform,
(ii) multiply by a filter and (iii) inverse Fourier transform. Convolution performs
this same operation in the time or space domain but, in general, is less efficient.
If, however, a filter can be represented in the time or space domain with
only a small number of components that are non-zero, then filter implementation
using convolution becomes more efficient. Many image processing filtering
operations are implemented by convolution. The non-zero portion of the time
414
or space representation of these filters is called a kernel. In this section, many

examples of small kernels are given, which are used in image processing.
The act of multiplying a signal by a function in the time or space domain
is sometimes called windowing. Multiplying a signal by a window changes the
amplitude of the time or space components of a signal. Owing to duality, filtering
and windowing are essentially the same operation, and the distinction is not
always maintained.
Psychologists refer to the just noticeable difference as the smallest difference
that can be perceived by humans. In vision, the just noticeable difference is a
function of the spatial frequency (more exactly the angular frequency). Humans
have limited spatial resolution and cannot detect variations at very high spatial
frequency. At somewhat lower frequency, detail can be perceived, but only if
it is very high contrast. As the frequency decreases, lower and lower contrast
variations can be perceived. At very low spatial frequencies, sensitivity to low
contrast again decreases. The just noticeable contrast sensitivity peaks at about
2 cycles per degree and falls off rapidly for higher and more slowly for lower
angular frequencies. These characteristics are important in image filtering design,
which often seeks to maximize transfer of the image information to a human.
12.3.4.1. One dimensional processing
A 1D signal f(t) can be filtered with a frequency domain signal H() by
transforming f(t) to F(), multiplying the signals to obtain G()=F()H() and
then inverse transforming G() to produce the result g(t). Alternately, the time
domain representation of H(), h(t), can be convolved with f(t):
g (t ) =
h(t ) f (t) dt (12.15)
A good example of 1D processing is a graphic equalizer. The sliders

on a graphic equalizer represent H(). Each slider corresponds to a range of
frequencies. The tones in the input signal are multiplied by the values represented
by the slider to produce the output signal.
12.3.4.2. Two- and three dimensional processing
Two dimensional processing is exactly analogous to 1D processing with
the 1D variables t and replaced with the 2D variables x, y and kx, ky. For 3D
processing, t is replaced by x, y, z, and w is replaced by kx, ky, kz. Four-, five-,
six-, etc. dimensional processing can be performed similarly.
415
CHAPTER 12
A property of the Fourier transform equations is separability the

transform with respect to x can be performed first followed by the transform
on y. It is sometimes convenient to implement a 2D transform by first doing the
transform on the rows, and then doing the transform on the columns. All that is
needed is a 1D Fourier transform subroutine, and any number of dimensions can
be implemented just by managing the indexing.
12.3.5. Band-pass filters
Band-pass filters maintain a range of frequencies while eliminating all
other frequencies. An ideal band-pass filter is exactly one for some range of
frequencies and exactly zero for the remaining frequencies. There is a very sharp
transition between the pass- and the stop-zones. A general heuristic is that a
sharp edge in one domain will tend to create ripples in the other domain. Ideal
band-pass filtering often creates ripples near sharp transitions in a signal. It is,
therefore, common to make the transition from the pass-zone to the stop-zone
more gradual.
12.3.5.1. Low-pass, noise suppression and smoothing filters
A low-pass filter is a type of band-pass filter that passes low frequencies and
stops high frequencies. In an image, the high spatial frequencies are needed for
detail. Zeroing the high spatial frequencies smooths the image. By suppressing
the high frequencies compared to the lower frequencies, the image noise is often
suppressed. Low-pass filters both smooth an image and decrease the noise in an
image.
Information theory tells us that image processing can only lower the
information in an image. (An exception is when processing includes a priori
information.) Noise suppression filtering lowers the information carried by the
noise. It is common to say that processing brings out a feature, but actually
what it is doing is suppressing the noise which is confusing to the human viewer.
All of the information about the content exists in the original image; processing
merely reduces the useless information contained in the noise.
A somewhat similar effect can often be obtained by simply moving away
from the image or by making an image smaller in size. Moving away makes the
high frequency noise correspond to angular frequencies where vision has low
contrast sensitivity. These operations are not entirely equivalent to low-pass
filtering. They may also move the important content to a point where contrast
sensitivity is reduced. Instead, it is best to view an image at a size where the
content is at the maximum visual sensitivity and the noise is suppressed with a
low-pass filter.
416
A common method of implementing a low-pass filter is as a convolution

with a small kernel. For example, Table12.3 shows a 33 smoothing kernel, also
commonly called a 9-point smoothing kernel. Each point in the smoothed image
is made up of the scaled sum of nine surrounding points. Also shown is a 55
smoothing kernel. Each point is made up of the scaled sum of 25 surrounding
points. The top of Fig.12.2 shows the effect of the 55 kernel on a circular
TABLE12.3. COMMONLY USED KERNELS IN IMAGE PROCESSING
Smoothing
33

1 2 1
2 4 2
1 2 1

Smoothing
55
1
2
4
2
1
2
4
8
4
2
3 2
8 4
16 8
8 4
3 2
Sharpening
1
1
1
1
1
1
1
1
1
1
1 1 1
1 1 1
25 1 1
1 1 1
1 1 1
Unsharp mask
0
0
1
0
0
0
1
2
1
0
1 0 0
2 1 0
17 2 2
2 1 0
1 0 0
X gradient
0
0
0
0
0
1
1
1
1
1
0 1
0 1
0 1
0 1
0 1
Y gradient
0
1
0
1
0
0
1
0
1
0
0 0 0
1 1 1
0 0 0
1 1 1
0 0 0
1
2
4
2
1
0
0
0
0
0
417
CHAPTER 12
region of constant intensity. The effects are relatively small and are limited to
the edges of the circle. The bottom of Fig.12.2 shows the effect on a very noisy
image of the same object. In this case, the effect is much more pronounced.
FIG.12.2. The upper left quadrant shows a circular region of constant intensity; the upper
right quadrant shows the effect of applying a 5-by-5 smoothing kernel; the lower left quadrant
shows a very noisy version of the circular region; the bottom right quadrant shows the effect of
applying a 5-by-5 smoothing kernel.
A common method of specifying a low-pass filter in the frequency domain

is given by the Butterworth filter:
H (k ) =
1
(1 + (k / k 0 ) 2 n )
(12.16)
The Butterworth filter has two parameters, k0 and n. The parameter k0 is the
cut-off frequency and n is the order. The filter reaches the value 1/ 2 when the
spatial frequency k is equal to k0. The parameter n determines the rapidity of the
transition between the pass-zone and the stop-zone.
The filter is sometimes shown (Fig.12.3) in terms of the square of the filter,
1/(1+(k/k0)2n). In that case, the filter reaches the value of half at the cut-off
frequency. Confusion can arise since sometimes the filter itself is defined as the
square value, i.e. without the square root. Furthermore, the filter is sometimes
defined with an exponent of n instead of 2n.
418
FIG.12.3. This figure shows the square of the Butterworth filter in the spatial frequency
domain with a cut-off equal to 0.25 cycles/pixel and n equal to 2, 3, 5, 7 and 10.
12.3.5.2. High-pass and edge-detection

Edges, and more generally detail, in an image are encoded with high spatial
frequencies. Large regions of constant intensity are coded with low spatial
frequencies. Thus, a high-pass filter will tend to emphasize the edges in an image.
A common method of implementing a high-pass filter is as a convolution
with a small kernel. Table12.3 shows a 5-by-5 sharpening kernel. The top of
Fig.12.4 shows the effect of applying this filter to a circular region with smooth
edges. A high-pass filter will also emphasize pixel to pixel noise. The bottom of
Fig.12.4 shows the same circular object with a small amount of noise. Applying
the high-pass filter to this image results in an image where the noise predominates.
In general, a high-pass, edge-detection filter is used when the starting image has
low noise.
419
CHAPTER 12
FIG.12.4. The upper left quadrant shows a smooth circular region; the upper right quadrant
shows the effect of applying the 5-by-5 sharpening kernel; the lower left quadrant shows a
circular region with a small amount of noise; the bottom right quadrant shows the effect of the
sharpening kernel on amplifying noise.
12.3.5.3. Unsharp masking and retinal ganglion cells

A band-pass filter, which passes intermediate frequencies and stops both
low and high frequencies, will both smooth and edge enhance. It will smooth
over a short range of pixels and it will enhance edges between regions at a
slightly longer distance. As the low-pass and high-pass filters, this filter is often
implemented in the space domain by convolution with a small kernel. Table12.3
shows a band-pass kernel. The pixels near the centre of the kernel will be added
together; those pixels will smooth the highest frequencies in the image. An
annulus of pixels at a greater distance is subtracted from the centre pixels. This
part of the kernel will emphasize edges that exist over this distance.
Two dimensional kernels with this form are sometimes called Mexican
hat kernels because the space of the kernel looks something like a Mexican hat.
A smooth Mexican hat filter can be made from the difference of two Gaussians.
This same operation is also called unsharp masking due to the similarity of a film
processing method of that name.
High latitude means that images with a large dynamic range, with bright
sunlight and shadows, can be displayed. Since film does not have the dynamic
range of the human eye, the only way to obtain the highest latitude is to have
low contrast. Even worse, a printed image, even a glossy image, has a dynamic
range many orders of magnitude lower than the human eye. In an unsharp
masked image, the intensity contrast over the range of the unsharp kernel is
recovered from a low contrast original. The contrast between more distant
420
objects is reduced. Although markedly altered, unsharp masked images often

look verisimilar to the natural scene. In fact, they sometimes look better than
the natural scene. Unsharp masking is used extensively in coffee table books.
In order to understand why these distorted images look verisimilar to the
original scene, it is useful to consider the first ganglion cells in the retina. The
first ganglion cells in the retina have a centre/surround organization. There is
a small region on the retina where the cells have an excitatory input to a cell.
There is a larger surrounding region where the cells have an inhibitory input to
a cell. What is coded in the optic nerve is the relation of the small excitatory
regions to the surrounding inhibitory regions. The ganglion cell processing is
very much like unsharp masking. The brain then decodes these inputs to make a
final impression of what the scene shows.
One important feature of human vision is its ability to discount illumination.
To a very large extent, humans see the same colours under a wide variety of
illuminations. The visual system discounts illumination. In the scene with bright
sunlight and shadows, the human is aware of the illumination, but within a region,
objects are still recognized even though the light reflecting from them may be
orders of magnitude different depending on what part of the scene they are in. It
is likely that this centre/surround mechanism is key for discounting illumination.
The important thing to remember in terms of image processing is that even
though an unsharp masked image may appear to humans as verisimilar to the
natural scene, it has been considerably altered from the original scene.
12.3.6. Deconvolution
Deconvolution is the process of undoing convolution. If we know the input,
f(t), to a system, h(t), and the output, g(t), from the system, then deconvolution
is the process of finding h(t) given f(t) and g(t). In the frequency or spatial
frequency domain, deconvolution is straightforward a simple division:
H ( ) =
G( )
(12.17)
F ( )
There is a problem with this simple deconvolution. If F() is zero for

any , then H() is infinite. Furthermore, since nuclear medicine deals with
real systems that have noise, any time F() is small, then any noise in G()
is amplified. The solution is to filter the simple deconvolution by a filter that
depends on the power in F().
Since signals usually have less power at high frequency or high spatial
frequency, the first practical deconvolution was performed by simple high-pass
filtering. A Metz filter (Section 12.3.7.2) can be used to emphasize the
421
CHAPTER 12
mid-frequency range while attenuating the high frequencies. If the statistical

properties of the signal and noise are known or can reasonably be approximated,
then a Wiener filter (Section 12.3.7.3) can be used.
12.3.7. Image restoration filters
Some filters attempt to restore degraded images. Degradations can come
from many sources, but the most important for this book is degradations caused
by imaging instruments such as gamma cameras. If the imaging system is linearshift-invariant, then it can be modelled by convolution. A restoration filter needs
to undo the effect of the imaging system, deconvolving for the system function.
Even in the case where the system is not linear or shift-invariant, it is often
possible to ameliorate the effects of a system with a filter.
12.3.7.1. Ramp
The effect of a projection/back projection is to smooth an object
(seeSection 12.3.5.1). In polar coordinates, the system function is given by the
simple equation:
H (k , ) =
1
(12.18)
k
It should be noted that the system function is not a function of . The higher
frequencies in the image are reduced by a factor proportional to their spatial
frequencies.
The obvious method of restoring the original image is to multiply it by a
restoration filter given by:
G(k , ) = k (12.19)
The restoration filter increases in magnitude in proportion to the spatial

frequency. Owing to the shape of the filter, it is called a ramp filter. A ramp filter
reverses the projection/back projection operation.
12.3.7.2. Metz
If the noise in the detected signal is uniform, then a restoration filter will
alter the noise spectrum. For small restorations, the signal to noise may still be
favourable, but for frequencies with large restorations, the noise may dominate.
The best result is often to restore small effects completely, but restore large
422
effects less. The Metz filter combines these two goals into a single filter. For any
system function, H(k, ), the Metz filter is given by:
G(k, ) = H 1 (k, )(1 (1 H (k, ) 2 ) x ) (12.20)
MTF.
The system function is often written as MTF(k, ) to emphasize that it is an
The first term in this equation, 1/H(k, ), reverses the effect of the system.
When H(k, ) is nearly one, the second term is about equal to one; when H(k, )
is nearly zero, the second term is about equal to zero; at intermediate values,
the second term transitions smoothly between these two values. In Fig.12.5, a
simulated system function is shown by the dotted lines. Four Metz filters with
different X parameters are shown. At low frequencies, the Metz filter counteracts
the effects of the imaging system. At high frequencies, it takes on the character
of a low-pass filter. The transition between characters is controlled by the
parameter X.
12.3.7.3. Wiener
Wiener filtering is used when the statistical properties of the signal and of
the noise are known. A function known as the power spectral density, S(), gives
the expected amount of power in a signal as a function of frequency. Power
means that the function is related to the square of the magnitude of a signal. The
Wiener filter is given by:
H ( ) =
S f ( )
S f ( ) + S n ( )
(12.21)
where Sf () is the power spectral density of the signal f(t), and Sn() is the
power spectral density of the noise n(t). Under the proper circumstances, it can
be shown that the Wiener filter is the optimal estimate of a process f(t) given the
noisy data, f(t) + n(t).
For those frequencies where the signal is much larger than the noise, the
Wiener filter is equal to one. For those frequencies where the noise is much larger
than the signal, the Wiener filter is equal to zero. The Wiener filter transitions
smoothly from the pass-zone to the stop-zone when there is more noise in the
data than signal. The Wiener filter provides a solid mathematical basis for the
heuristic that has already been used several times.
423
CHAPTER 12
FIG.12.5. The dotted line shows a simulated modulation transfer function (MTF). The four
solid lines show Metz filters for X equal to 1.5, 2, 2.5 and 3.
12.3.8. Other processing

This chapter is predominantly concerned with linear-time-invariant or
linear-shift-invariant signals. Fourier transform methods can be applied to these
signals. Fourier transforms provide considerable insight and using the fast Fourier
transform algorithm, they are very efficient computationally. However, the types
of system that can be modelled are limited. Any real imaging system typically
has an FOV. At the edge of the FOV, the system is no longer shift-invariant.
Furthermore, attenuation, scatter and depth dependent resolution cannot be
modelled using linear-shift-invariant systems.
424
12.3.8.1. Linear systems

Linear systems without being limited to time- or shift-invariance are much
more powerful. Limited FOV, attenuation, scatter and depth dependant resolution
can all be modelled. In addition, non-standard imaging geometries such as limited
angle tomography can be modelled.
The relation between the input and the output of a linear system is
equivalent to a system of simultaneous equations. An efficient method of writing
a set of simultaneous equations is to use matrix algebra:
g = H f (12.22)
where f and g are vectors and H is a matrix.

In the case of an image or especially a volume, the vectors f and g represent
the whole image or volume. In the case of a 256256256 volume, each vector
has 16 mega elements. The matrix, which is 16 mega elements by 16 mega
elements, has 256 tera elements. It should be obvious that this simple expression
hides a great deal of complexity. However, it does bring out the analogy between
linear systems and linear-time-invariant or linear-shift-invariant systems.
12.3.8.2. Non-linear systems
The output of a linear system to the sum of two signals is equal to the sum
of the outputs from each signal separately. Furthermore, the output of a system to
a signal multiplied by a number is equal to the output due to the signal multiplied
by the same number. Real systems are almost never linear. There is usually some
size input for which real systems break down. Above that magnitude, increases in
the input are not reflected in the output.
Non-linear systems are very difficult to deal with. Consequently, non-linear
systems are often modelled by restricting their application to the range where
the system is nearly linear. Other times, it is possible to transform the input or
the output in order to make the system behave approximately as a linear system.
There are methods of dealing with non-linear systems, but since they are usually
very difficult, non-linear systems are rarely modelled directly.
12.4. DATA ACQUISITION
Modern nuclear medicine data acquisition devices have considerable
computing power embedded in them that performs many tasks to improve
425
CHAPTER 12
images. These interesting and important computing functions are described in

Chapter11. This chapter will treat the camera as a black box and begin with the
data that have already been obtained.
12.4.1. Acquisition matrix size and spatial resolution
For planar imaging, the FOV of a camera is divided into discrete
elements that correspond to the pixels in a digital image. Most commonly,
a rectilinear system is used and the digital image is analogous to a matrix
(seeSection 12.2.3.2). The size of the FOV and the spatial resolution of the
camera are important in choosing the size of the matrix to be used. If the imaging
device is modelled as a linear space-invariant system, then to avoid aliasing, the
sampling should be twice the highest frequency that can be recorded by a camera
(seeSection 12.3.2). For example, if the FOV of the camera is 40cm and the
camera can detect frequencies up to 2 cycles/cm, then the matrix should have:
2 samples/cycle2 cycles/cm40cm=160 samples
For a normal large FOV Anger camera, a 256256 matrix is more
than sufficient for most imaging situations. For clinical imaging, a 128128
matrix is often sufficient. For rapid dynamic images that are often very count
limited, a 6464 matrix can be used. For vertex to thigh imaging, a 1:4 matrix,
e.g. 2561024, will cover 160cm. When including the legs, a non-power-of-two
matrix may be most logical.
In the past, when memory and storage were more expensive, there was
emphasis on using as small a matrix as possible. Currently, it makes more sense
to emphasize that the use of a matrix is large enough to be sure that imaging is
not limited.
12.4.2. Static and dynamic planar acquisition
Many of the processes of interest in nuclear medicine involve a changing
pattern of distribution of a radiopharmaceutical in the body. In matrix-mode, a
sequence of images collected over time allows visualization and measurement
of the biodistribution of the radiopharmaceutical. Each image is often called a
frame of data. Collection of a sequence of images over time is called a dynamic
acquisition; collection of an image at a single point in time is called a static
acquisition. Several static images may be grouped together as a multiple static
acquisition; however, the difference between multiple static and dynamic
acquisitions is largely conceptual. A dynamic acquisition that shows vascular
426
delivery of the radiopharmaceutical is often called a flow study. A dynamic

collection during renal scintigraphy is called a renogram.
The FOV of the gamma camera is divided into discrete rectangles that are
generally square. Each rectangle corresponds to one pixel in the image and each
is represented by one memory location. The size of the rectangles puts a limit
on the resolution that is achievable, but the rectangle size is not the resolution.
The resolution of a gamma camera is the overall ability to resolve adjacent
sources of activity. For each scintillation, the gamma camera identifies the
rectangle corresponding to the location of the scintillation and increments the
corresponding memory location.
12.4.3. SPECT
SPECT is typically performed by acquiring a sequence of 2D projection
images from multiple different angles around the object. The 2D projections can
be represented by p[x, z], where z is the axial position. The set of projections can
be described by a 3D function, p(x, z, ), where is the position of the camera
head.
12.4.3.1. Sinogram
The first step in reconstruction is to reformat the data in terms of axial
position. The reformatted data can be described by a function p(x, , z). For
each axial position, the set of data, p(x, ), has one line of data from each of
the projections. Each of the images, p(x, ), is called a sinogram. The x position
in the sinogram is the same x position in the raw projection image. The other
position is the projection angle. A point source will project onto all of the
projection images at the same axial position z. The position of the point source in
the direction will form a sine wave in the sinogram, hence the name sinogram.
The sinogram is not a particularly useful way of visualizing the data, but it
is frequently available in nuclear medicine systems. It allows easy identification
of a few artefacts. If there is patient movement during data acquisition, there will
be discontinuity of the sine waves at the projection angle where the movement
occurred. A detector problem will show up as straight lines of different intensity
in the direction. Intermittent problems with the whole detector will result in
straight lines in the x direction.
12.4.3.2. Sampling requirements in SPECT
Each of the sinograms p(x, ) needs to be reconstructed into an image
f(x, y) where x and y are coordinates with respect to the patient. The two types
427
CHAPTER 12
of reconstruction are Fourier reconstruction and iterative reconstruction. For

an adequately sampled linear-shift-invariant system, Fourier reconstruction is
exact in the absence of noise. Iterative reconstruction is approximate, but allows
a much more realistic model of the system including attenuation, scatter, depth
dependent collimator resolution and non-regular sampling.
A good heuristic is that in order to reconstruct N points in a slice,
N data samples are needed. Sampling tangentially to the slice is reasonably
straightforward; x should have the same sampling rate as x and y. A good rule for
Fourier reconstruction is that the number of angular samples over 180 is n/2,
where n is the samples along x. This number is equal to the heuristic within a
factor of /2.
Resolution recovery is used in several nuclear medicine applications.
Resolution recovery improves resolution compared to the heuristic given above.
The extra information required to improve resolution is supplied by an a priori
assumption, e.g. the edges in the image are relatively sharp. If the object does
not meet the a priori assumption, then artefacts will be produced. As long as the
assumptions are reasonable, resolution recovery can be useful.
Well designed collimators always involve a trade-off between resolution
and sensitivity. A good heuristic is that for optimal detection, the collimator
should have a resolution equal to the object to be detected. However, there
may be more than one resolution object to be detected. For example, in cardiac
imaging, if a low resolution initial image is needed for positioning followed by
a high resolution image, then using resolution recovery for the higher resolution
image can ameliorate use of a lower resolution collimator.
Each tomographic data sample includes a combination of the voxels in the
object. For detectors that measure energy, so-called square-law detectors, the
data samples will always be sums of object voxels. (This limitation is not true
for MRI where the data have both an amplitude and a phase.) For tomographic
data, square-law detectors always oversample low frequencies and reconstruction
must amplify the high frequencies compared to the low frequencies. Typically,
noise is relatively uniform in the projection space. Thus, after reconstruction, the
high frequency noise is amplified with respect to the low frequency noise.
12.4.4. PET acquisition
The detectors in positron cameras detect single photons. However, the key
feature of a positron camera is detection of pairs of photons in coincidence. Two
photons detected in coincidence define an LOR. The LORs are then reconstructed
into a 3D image set. At that point, PET and SPECT data are the same.
Most PET cameras are made from individual crystals. The singles data
from the crystals can be recorded in a number of memory elements equal to the
428
number N of crystals. NN memory elements are needed for the LORs. (Some
potential LORs do not traverse the imaging area.) The crystals in a PET camera
make a 2D array. In the usual cylindrical geometry, the two dimensions are the
axial slice and the angular position around the ring of detectors for that slice.
LORs are defined by two crystals and four dimensions. To limit the electronics,
early PET cameras limited coincidences to a single slice. In that case, the slice
position is the same for both detectors, and the LORs are 3D, one axial and two
angular position dimensions.
12.4.4.1. 2D and 3D volume acquisition
Modern PET cameras can collect oblique LORs, greatly increasing
sensitivity. Some cameras have axial collimators that can be inserted or retracted
while others do not. Imaging with axial collimation involves a 3D object, a
3D image and 3D data. Strangely, this type of imaging is called 2D. Imaging
without axial collimation again involves a 3D object and a 3D image, but in
this case the data are 4D. This type of imaging is called 3D imaging. Although
these terms for imaging with and without axial collimation do not make sense,
they are imbedded in the PET vernacular.
12.4.4.2. Time of flight
Time of flight (TOF) imaging measures the difference in arrival of the two
photons at the two detectors. The difference in arrival time can be used to position
the annihilation event along the LOR. This provides an additional dimension to
the data. Non-axial collimation, TOF PET data are 5D. The resolution along the
LOR is relatively low. The speed of light is about 3108 m/s, which means
that in 1 ns photons travel 30cm. Current detectors can detect differences in
arrival of about 0.5 ns. Thus, improvements using TOF detection are important,
but modest.
12.4.4.3. Sampling requirements in PET
Data sampling is determined by the LORs. Determination of the resolution
of a PET camera is complicated, but once it is determined, then the size of the 3D
reconstructed matrix is the same as in SPECT imaging. Non-axial collimation
PET adds oblique data with the potential of increasing the axial dimension.
In practice, the axial resolution is similar for axial collimation and non-axial
collimation PET. With cylindrical cameras, there are many more oblique angles
that can be detected on the centre slice than on end slices. In non-axial collimated
PET, the end slices are noisy. The non-uniform axial sampling is ameliorated in
429
CHAPTER 12
whole body imaging by overlapping the different bed positions, so that the data
are more uniformly sampled.
12.4.5. Gated acquisition
Data acquisition can be gated to a physiological signal. Owing to count
limitation, if the signal is repetitive, the data from multiple cycles are usually
summed together. A gated, static acquisition will have three dimensions
two spatial and one physiological. A gated dynamic acquisition will have four
dimensions two spatial, one time and one physiological. A gated SPECT
or PET acquisition will have a different four dimensions three spatial and
one physiological. A gated dynamic SPECT or PET acquisition will have five
dimensions three spatial, one time and one physiological. Count limitations
will often limit the dimension and/or the elements per dimension.
12.4.5.1. Cardiac-synchronized
Gated cardiac studies were one of the early studies in nuclear medicine.
Usually, the electrocardiogram is used to identify cardiac timing. The
R-wave is relatively easy to detect; however, at least a little knowledge of
electrocardiography is useful to know how to select electrode positions where the
R-wave has a high amplitude.
The timing of events during the cardiac cycle changes with cycle length.
The timing of contraction, systole, and the initial part of relaxation, diastole,
change relatively little as cycle length changes. Most of the change in cycle
length is a change in diastasis, the period between early relaxation and atrial
contraction. Atrial contraction, atrial systole, has a relatively constant relation
to the end of the cardiac cycle, i.e. to the next R-wave. Thus, for evaluation of
systolic and early diastolic events, it is better to sum cycles using constant timing
from the R-wave than to divide different cycles using a constant proportion of the
cycle length.
However, with constant timing, different amounts of data are collected in
later frames depending on the number of cycles that are long enough to reach
those frames. Normalizing the data for the acquisition time for each frame can
ameliorate this problem. Alternatively, systole and early diastole can be gated
forwards in time from the preceding R-wave, and atrial systole can be gated
backwards in time, joining the two during diastasis.
430
12.4.5.2. Respiratory-synchronized
Gating to respiration has been used in nuclear medicine both to study
respiratory processes and to ameliorate the blurring caused by breathing.
The signal to use for respiratory synchronization is not clear-cut. Successful
synchronization has been based on different types of plethysmography, on
chest or abdominal position changes, and on image data. Plethysmography,
measurement of the air in the lungs, can be measured by integrating the rate of
airflow or indirectly from changes in temperature of the air as it is breathed in
and out. Corrections often need to be made for errors that build up from cycle
to cycle. Electrical impedance across the chest changes with respiration and can
also be used to detect lung volume changes. Chest and abdominal motion often
correlate with the respiratory cycle, although detection of motion and changes
in diaphragmatic versus chest wall breathing also pose problems. Although
difficult to measure, respiratory gating is becoming more common, particularly
in quantitative and high resolution PET applications.
12.4.6. List-mode
Instead of collecting data as a sequence of matrixes or frames, the position
of each scintillation can be recorded sequentially in a list. Periodically, a special
mark is entered into the list that gives the time and may include physiological
data such as the electrocardiogram or respiratory phase.
List-mode data are generally much less efficient than matrix-mode data.
For example, a 1 million-count frame of data collected into a 256 by 256 matrix
will require 64000 memory locations. In list-mode, the same data will require
1 million memory locations. The reason list-mode is so much less efficient is
that it contains extra information, namely the order of arrival of each of the
scintillations. However, this information is not interesting. List-mode is more
efficient than matrix-mode when there is less than one event per pixel on average.
Since an image with one event per pixel is very noisy, it may be surprising
that list-mode is useful at all. However, with gated images, multiple individual
images are added to produce one output image. List-mode can be more efficient
if it is necessary to keep the separate images that make up one gated image for
post-processing.
12.5. FILE FORMAT
The nicest thing about standards is that there are so many of them to
choose from. Ken Olsen, founder of Digital Equipment Corp.
431
CHAPTER 12
12.5.1. File format design

12.5.1.1. Interfaces
Standards have facilitated the rapid development of computers. Standards
can be considered as part of a larger topic, interfaces. In general, an interface is
a connection between an entity and the rest of the system. If each entity in the
system has an interface and if the interface is the only allowed way for the other
parts of the system to interact with an entity, then there is a great simplification
of the overall system. Development of one part of a system depends only on
that part and the interface. It does not depend on any other part of the system.
Each part is greatly simplified because it is separate from the rest of the system.
A system that is divided up into parts with well designed interfaces between the
parts is called modular.
There was an early problem with development of complex systems. As
more resources such as programmers were added to a task, very little additional
output resulted. The problem was that as the task grew in complexity, more
time was spent on communication between the programmers and less was spent
on programming. At times, the output actually decreased as more resources
were devoted to the task. The problem is that as a project becomes larger, the
complexity tends to increase exponentially.
In a modular system, the details of each portion of the task are hidden from
other parts of the task. Data hiding is a goal for good system design. Modularizing
a task tends to linearize the complexity. As a task with complexity n becomes
twice as large, the work becomes about 2n, not n2 or e2n times as much.
File formats are a type of interface. They define how information will be
transferred from one system to another. By having a well designed format, each
system becomes separate. The goal of this section is to describe existing file
formats and understand how they can simplify the design of a nuclear medicine
information system.
12.5.1.2. Raster graphics
There are two general image formats raster graphics and vector
graphics. Vector graphics define an image in terms of components such as points,
lines, curves and regions, e.g. polygons. Points, lines, curves and the boundaries
of regions can have a thickness and a colour. The interiors of regions can have
colours, gradients or patterns. For images that are made up of these types of
component, the vector graphics description is very compact. Vector graphics
can be smoothly expanded to any size. Modern type faces, which are defined
in terms of vector graphics, can be expanded to any size while retaining their
432
smooth outlines. Vector graphics are used extensively in gaming software. Vector
graphics are not useful for nuclear medicine images, so this chapter will only
describe raster graphics.
Raster graphics images are made up of a matrix of picture elements or
pixels. Each pixel represents one small rectangle in the image. The primary data
in nuclear medicine are numbers of counts. Counts are translated into a shade of
grey or a colour for display. There are many ways the shade of grey or the colour
of a pixel can be encoded. One factor is the number of bits that are used for
each pixel. Grey scale images often have 8 bits or 1 byte per pixel. That allows
256 shades of grey, with values from 0 to 255.
A common way to encode colour images is in terms of their red, green and
blue (RGB) components. If each of the colours is encoded with 8 bits, 24 bits
or 3 bytes are needed per pixel. If each colour is encoded with 10 bits, 30 bits
are needed per pixel. RGB encoding is typical in nuclear medicine, but there
are other common encodings such as intensity, hue and saturation. For printing,
images need to be encoded in terms of cyan, magenta and yellow. Use of more
complicated encodings allows more vibrant images that use additional colours,
e.g. black, orange, green, silver or gold.
12.5.1.2.1. Transparency
Graphic user interfaces often allow a composition of images where
background images can be seen through a foreground image. In such cases, each
pixel in an image may be given a transparency value that determines how much
of the background will come through the foreground. A common format is 8 bits
for each of the RGB colours and an 8-bit transparency value, giving a total of 32
bits per pixel.
12.5.1.2.2. Indexed colour
Typically, an image will include only a small number of the 16 million (224)
possible colours in a 24 bit RGB palette. A common method of taking advantage
of this is to use indexed colour. Instead of storing the colour in each pixel, an
index value is stored. Typically, the index is 8 bits, allowing 256 colours to be
specified. In the case where more than 256 colours are used, it is often possible
to approximate the colour spectrum using a subset of colours. Sometimes, a
combination of colours in adjacent pixels will help to approximate a broader
spectrum. The algorithms for converting full spectrum RGB images into a limited
palette are remarkably good. It often requires very careful inspection of a zoomed
portion of the image to identify any difference.
433
CHAPTER 12
For indexed colour, a colour palette is stored with the images. The colour
palette has 256 colour values corresponding to the colours in the image. The
8 bit index value stored in a pixel is used to locate the actual 24 bit colour in
the colour palette, and that colour is displayed in the pixel. Each pixel requires
only 8 bits to store the index as opposed to 24 bits to store the actual colour. The
colour palette introduces an overhead of 8 bits for each of the 256 colours, but
since most images have tens of thousands or hundreds of thousands of pixels this
overhead is small.
12.5.1.2.3. Compression
The information content of an image is often much smaller than the
information capacity of an image format. For example, images often have large
blank areas or areas that all have the same colour value. Therefore, it is possible
to encode the image using less space. Improving efficiency for one class of
images results in decreasing efficiency for another type of image. The trick is to
pick an encoding which is a good match for the range of images that are typical
of a particular application.
One of the simplest and easiest encodings to understand is run-length
encoding. This method works well for images where large areas all have the
same value, e.g. logos. Instead of listing each pixel value, a value and the number
of times it is repeated are listed. If there are 50 yellow pixels in a row, rather than
listing the value for yellow 50 times, the value 50 is followed by the value for
yellow. Two values instead of 50 values need to be listed.
Another common encoding is called LempelZivWelch (LZW) after its
developers. It was originally under patent, but the patent expired on 20 June 2003,
and it may now be used freely. The LZW algorithm is relatively simple but
performs well on a surprisingly large class of data. It works well on most images
and works particularly well on images with low information content, logos and
line drawings. It is used in several of the file formats described below.
Both run-length encoding and LZW encoding are non-destructive or
reversible or lossless coding; the original image can be exactly reconstructed
from the encoded image. After destructive or irreversible or lossy coding, the
original image cannot be exactly recovered from the coded image. However,
much more efficient coding can be performed with destructive encoding. When
non-destructive encoding results in reduction of the image size by a factor of 23,
destructive encoding will often result in a reduction of the image size by a factor
of 1525. The trick is to pick an encoding system where the artefacts introduced
by the coding are relativelyminor.
The human visual system has decreased sensitivity to low contrast, very
high resolution variations. Details, high resolution variations, are conspicuous
434
only at high contrast. Discrete cosine transform (DCT) encoding takes advantage
of this property of the visual system. DCT encoding is in the Joint Photographic
Experts Group (JPEG) standard. The low spatial resolution content of the image
is encoded with high fidelity, and the high spatial resolution content is encoded
with low fidelity. The artefacts introduced tend to be low contrast detail. High
contrast detail, and both low and high contrast features at low resolution are
faithfully reproduced. Although there are artefacts introduced by the coding, they
are relatively inconspicuous for natural scenes. For nuclear medicine images, the
artefacts are more apparent on blown up images of text.
Wavelets are a generalized form of sinusoids. Some improvement in
compression ratios at the same level of apparent noise can be obtained using
wavelet-transform coding. The blocky artefacts that degrade DCT images are
not seen with wavelet-transform images. The JPEG 2000 standard uses wavelettransform coding.
Non-destructive coding systems often make use of the fact that adjacent
pixels are equal. The statistical noise in nuclear medicine images reduces the
similarity of adjacent pixels, thus reducing the utility of non-destructive coding.
Destructive coding may overcome this limitation, and since the statistical
variations generally do not carry any significant information, image quality may
not be greatly degraded.
12.5.2. Common image file formats
Common image file formats could be used for some or all of the raw nuclear
medicine image data. In fact, they are rarely used for this purpose. However,
secondary images, especially when used for distribution of image information,
generally use these standard file formats. This section will sketch the format of
these files, the advantages and disadvantages of the formats, and the typical uses
of these formats.
12.5.2.1. Bitmap
Bitmap (BMP) is a general term that may refer to a number of different
types of data. When used in reference to image formats, it may be used to mean
a raster graphic as opposed to a vector graphic format, but it often refers to a
Windows image format. The Windows file format is actually called a device
independent bitmap (DIB). The external DIB file format is distinguished from
various device dependent internal Windows BMPs. File name extensions .bmp
and .dib are used for the BMP image file format. BMP may be used to imply an
uncompressed format, but the DIB format defines several types of compression.
435
CHAPTER 12
12.5.2.2. Tagged Image File Format

Aldus Corp. created the Tagged Image File Format (TIFF). Adobe Systems
Inc., which merged with Aldus, now owns the copyright. The TIFF format can
be used free of licensing fees. The TIFF format includes many types of image.
Most commonly, it is used for 8 bit grey scale, 8 bit indexed colour or 24 bit RBG
colour images. Images are typically compressed with the non-destructive LZW
algorithm.
The TIFF format is often used for high quality single images that are
non-destructively compressed. Although multiple other uses are possible, this
application is often thought of as the strength of the TIFF format. The wide
variety of options provided by the TIFF format is both a strength and a weakness.
Few programs support all of these options, so that a TIFF image produced by
one program may not be readable by another program. A particular weakness for
nuclear medicine data is that the multiframe option is rarely supported.
12.5.2.3. Graphics Interchange Format
CompuServe, now a subsidiary of AOL, developed Graphics Interchange
Format (GIF). GIF is a relatively simple indexed format with up to 8 bits per
element. Up to 256 colours are selected from a 24 bit RGB palette. Pixels
can be transparent, in which case the background colour is displayed. The
89a specification included multiple images that may be displayed as a cine.
Compression is performed with the LZW algorithm.
As the GIF format is relatively simple yet quite useful, it is very commonly
supported. All web browsers support this format. It is particularly good for
storing low information content pictures such as logos and clip art. It is also good
for images such as nuclear medicine images. Cine capability is widely supported,
so it is very convenient for showing dynamic, gated and 3D data.
12.5.2.4. Joint Photographic Experts Group
The Joint Photographic Experts Group developed the JPEG format in 1992.
JPEG was approved in 1994 by the International Organization for Standardization
as ISO 10918-1. The JPEG standard leaves some issues unspecified. The
JPEG File Interchange Format (JFIF) clarifies these issues. To indicate that
a JPEG image also follows this standard, it is sometimes called a JPEG/JFIF
image, but generally JFIF is assumed. The JPEG standard is a reasonably
simple, non-indexed grey scale and colour image format that allows adjustable
destructive compression.
436
The JPEG coding tries to match the coding to human vision. It uses more
precision for brightness than for hue. It uses more precision for low frequency
data than for high frequency detail. The JPEG format is particularly good at
compressing images of natural scenes, the type of images in routine photography.
All web browsers support this format, and it is very widely used in general
photography products. It is not a particularly good format for line drawings and
logos; the GIF format is better for these types of image.
12.5.3. Movie formats
Multitrack movie formats allow audio and video to be blended together.
Often, there are many audio tracks and many video tracks where the actual movie
is a blending of these sources of information. A key part of a multitrack format
is a timing track, which has information about how the tracks are sequenced and
blended in a final presentation. However, in nuclear medicine, there is rarely a
need for all of this capacity. Often, all that is needed is a simple sequence of
images. A more complex movie format can be used for this type of data, but often
a simpler format is easier to implement.
12.5.3.1. Image sequence formats
Of the image formats described so far, BMP, TIFF and GIF can be used
to define an image sequence. An extension of the JPEG format, JPEG 2000,
also allows image sequences. As with the multiframe version of BMP and
TIFF formats, JPEG 2000 is relatively poorly supported. However, the multiframe
version, 89a, of the GIF format is widely supported. It is supported by all web
browsers and by almost all image processing and display programs. The GIF
format is a logical choice for distribution of cine images.
12.5.3.2. Multitrack formats
There are several multitrack movie formats available. As newer formats
that are still under development, they tend to be controlled by particular vendors.
The AVI format is controlled by Microsoft, the QuickTime format is controlled
by Apple, the RealVideo format is controlled by RealNetworks and the Flash
format is controlled by Adobe.
12.5.4. Nuclear medicine data requirements
There are two types of information in a nuclear medicine study image
information and non-image information. As described in the previous section,
437
CHAPTER 12
there are several general purpose image and movie formats. Often, these formats
lack capabilities that would be optimal for medical imaging. However, some
formats, e.g. TIFF, are general enough to be used for the image portion of the
study data. The advantage of using a widely accepted format is that it allows the
great diversity of software available for general imaging to be adapted to medical
imaging.
12.5.4.1. Non-image information
There is unique nuclear medicine information that must be carried along
reliably with the images. This information includes: identification data, e.g. name,
medical record number (MRN); study data, e.g. type of study, pharmaceutical;
how the image was acquired, e.g. study date, view; etc. This information is
sometimes called meta-information. Most general image formats are not flexible
enough to carry this information along with the image.
12.5.4.1.1. American Standard Code for Information Interchange
Text information is usually and most efficiently coded in terms of character
codes. Each character, including punctuation and spacing, is coded as a number
of bits. Initially, 7 bits were used; 7 bits allowed 27=128 codes, which were
enough codes for the 26 small and 26 capital letters, plus a fairly large number
of symbols and control codes. Using 1 byte (8 bits) for each character meant that
the extra bit could be used for error checking using a scheme called parity. Error
checking each character became more of an issue, so the American Standard
Code for Information Interchange (ASCII) was extended to 8 bits, allowing the
addition of 128 new codes. Characters in many of the Latin languages could be
encoded using 8 bit ASCII.
12.5.4.1.2. Unicode
Computer usage has transcended national boundaries; it is now just as easy
to communicate with the other side of the Earth, as to communicate within a
single building. Internationalization has meant that a single server or client needs
to be multilingual. Multilingual programming is now common. The ASCII code
is not adequate for this task, so a multilingual code, Unicode, has superseded it.
For example, the Java programming language specifies that programs be written
in Unicode. Unicode uses 32 bits, and there are more than 100000 character
codes, including all common languages and many uncommon languages.
There are different ways to implement Unicode called, Unicode
Transformation Format (UTF) encodings. One of the most common is a system
438
called UTF-8, which uses a variable length coding system. Different numbers of
bytes from one to four are used for different character codes. If the first byte in
a code has a zero in the highest bit, then 1 byte is used. If the first byte in a code
has a one in the highest bit, then subsequent bytes are used. The single byte codes
are identical to the 7-bit ASCII codes. For a document that only includes the
characters in the 7-bit ASCII system, encoding in ASCII and UTF-8 are identical.
This results in an efficient encoding of Latin languages. In addition, any other
Unicode character Greek, Japanese or symbol can be included occasionally
in a predominantly Latin text while maintaining average coding efficiency.
12.5.4.1.3. Markup language
The difference between text editors and word processors is that the former
largely edit the characters, while the latter define how the document will appear
the size of the text, the spacing, indentations, etc. For a word processor, the
layout of the document as it is entered is the same as the layout when it is printed.
In computer jargon, what you see is what you get (WYSIWYG).
Markup has the advantage that it can be read by humans and can be edited
with any text editor. Hypertext Markup Language (HTML), the language used
by the World Wide Web, was originally a markup language. When HTML was
first developed, text editors were used to writing it. Page layout capabilities were
added, and now WYSIWYG editors are generally used. A markup language
allows other types of information to be included. For example, one of the key
features of HTML is that it includes hyperlinks to other HTML pages on the
Internet.
12.5.4.1.4. Extensible Markup Language
Currently, a very popular and increasingly used method for encoding text
information is a standard called Extensible Markup Language (XML). It provides
a method of producing machine-readable textual data. For example, the Real
Simple Syndication (RSS) standard is encoded with XML. A common use of the
RSS standard is for web publishing. Almost all newspaper web sites use RSS to
communicate with subscribers.
XML is not a markup language itself, but rather a general format for
defining markup languages. It defines how the markup is written. A document
is said to be well formed if it follows the XML standard. If it is well formed,
then the markup can be separated from the rest of the text. More information,
provided by an XML schema, defines a particular markup language. The most
recent version of HTML, XHTML, is fully compatible with the XML standard.
439
CHAPTER 12
For nuclear medicine file formats, one of the key properties of XML is
that it can be used to make text information readable by machine. For example,
consider the following section of an XML document:
<patient>
<name>
<last>Parker</last>
<first>Tony</fist>
</name>
<medical_record_number>10256892</medical_record_number>
</patient>
It would be straightforward for a computer to unambiguously determine
the name and MRN from this document. Human readability and text editor
processing of XML make it highly self-documenting. Owing to its wide
acceptance in the computer world, XML is the most appropriate current format
for storing non-imaging nuclear medicine information.
XML can be used to store numerical data as characters,
e.g. <number>12.56</number>. The scalable vector graphics format for vector
graphics is an XML language. However, XML is too inefficient for coding raster
graphics image information.
12.5.4.2. Non-image information in general formats
(a) JPEG 2000: An extension of the JPEG format, JPEG 20001, has several
very interesting capabilities. The JPEG 2000 standard is a general purpose
image format; however, it was developed with medical imaging as a
potential application. Since it uses wavelet-transform compression, the
image quality at high compression ratios is considerably better than with
DCT compression. JPEG 2000 allows multiple frames, so it could be used
for cine; it has an advantage over GIF of providing good compression
of information rich images such as natural scenes. However, the reason
that it is included in this section is its capability to carry considerable
meta-information in tight association with the image data. Although this
format was defined nearly a decade ago, it has not found wide acceptance.
The JPEG 2000 wavelet-transform coding is allowed in the Digital Imaging
and Communications in Medicine (DICOM) standard, but mass market
software, such as browsers, generally do not support this standard. It has
440
http://www.jpeg.org/jpeg2000/
been included not as a solution that is currently practical, but rather as

an example of a non-medical standard that has capabilities that could be
applied to the particular requirements of nuclear medicine.
(b) Portable Document Format (PDF): PDF has not been used extensively in
medicine, but it has many useful capabilities. It is intended for page layout.
It can contain text, audio, images and movies. It is very widely supported,
including support by all web browsers. Although uncommonly used in
medicine, it has many properties that would make it an excellent format for
distribution of image or cine information.
12.5.4.3. Image information
The image portion of a nuclear medicine study can be a sequence of
static images, a dynamic series, a gated series, multiple dynamics series, raw
data from a tomographic collection, reconstructed images, a dynamic series of
tomographic collections, a dynamic series of gated tomographic collections,
related tomographic datasets, curves from regions of interest, functional datasets
derived from calculations on other datasets, etc. Thus, it should be clear that a
rather general data format is needed. This section will present suggestions for
general image format principles.
12.5.4.3.1. Types of data
The types of data in the last paragraph are not only long, but also
incomplete; it is easy to think of other variations. A different data format for each
type of data would add complexity without functionality. Therefore, it is logical
to separate the interpretation of the data from the data. Whether the data represent
a dynamic series or a set of static planar images does not have to be represented
in the dataset. There is no necessity to have two dataset formats one for a
128128 image and another for a set of 128 curves with 128 points.
The interpretation of the data should be part of the non-image information,
rather than part of the data. The data type should be determined by the format
of the data elements. Separation of the interpretation from the data format will
simplify processing and handling of the data. Adding two 128128 images and
adding two sets of 128 curves with 128 points is exactly the same operation; an
add function should not concern itself with the interpretation of the data.
12.5.4.3.2. Data element
Several different data elements are needed. The raw data collection bins can
generally be 8 or 16 bit unsigned integers, depending on whether a maximum of
441
CHAPTER 12
255 or 65535 counts per pixel are required. Since processed data may require a
larger dynamic range, floating point or complex data may be appropriate in many
circumstances. For some analysis, signed integers may be most appropriate.
A region of interest can be represented as a single bit raster or with vector
graphics.
The selection of a small number of data element formats would simplify the
programming task. The trick is to limit the number of formats while maintaining
all of the functionality. It would be logical to select some standard set of data
element formats, especially a set of formats associated with a widely accepted
data standard. However, it is not clear that such a standard exists. At aminimum,
16 bit integer and floating point formats are required. Probably signed and
unsigned 8 and 16 bit formats should be included. It is not clear whether
increasing complexity by including a vector graphics format is worth the added
functionality.
12.5.4.3.3. Organization
A logical first level of organization of the image data is what will be called
a dataset. A dataset is an n dimensional set of data in which each of the data
elements is the same format, e.g. 8 bit unsigned integer, 16 bit signed integer
and IEEE 32 bit floating point data. The key characteristic is that the dataset is
a sequence of identical elements. The dimensions do not need to be the same;
7 sets of 100 curves with 256 points is a 7 100256 dataset.
The dataset should be the atom of the data; there should not be a lower
level of organization. The lower levels of organization will depend on the type of
data and the dimensions. For example, a 256256256 tomographic volume
could be considered as 256 axial slices of 256256, but it would be equally
valid to consider that dataset as 256 coronal slices of 256256. The organization
of a dataset consisting of 4D PET lines of response will depend entirely on
the reconstruction algorithm. The lower levels of organization depend on the
non-image information and should not be part of the image data format.
12.5.5. Common nuclear medicine data storage formats
12.5.5.1. Interfile
The Interfile format has been used predominantly in nuclear medicine.
The final version of Interfile, version 3, was defined in 1992. Although Interfile
has been largely replaced by DICOM, it has some interesting properties. The
metadata, encoded in ASCII, are readable and editable by any text editor. The
lexical structure of the metadata was well defined, so that it is computer readable.
442
12.5.5.2. Digital Imaging and Communications in Medicine

DICOM is the most widely accepted radiological imaging format.
DICOM began as ACR-NEMA, a collaboration between the American
College of Radiology and the National Electrical Manufacturers Association.
Version 3 of that standard changed the name to DICOM, in part to position
the standard in a more international framework. DICOM is often thought of
as a file format; however, the standard covers communication more broadly. It
defines a transmission protocol, a query and retrieval standard, and workflow
management. Unfortunately, it is overly complex, non-self-describing and has a
heavy 2D image bias (seeSection 12.6.4).
12.6. INFORMATION SYSTEM
12.6.1. Database
Databases are one of the most common applications for computers. Web
sites frequently depend heavily on a database. For example, search engines
provide information from a database about the web and on-line retailers make
extensive use of databases both for product searches and for information about
customers and orders.
12.6.1.1. Table
Almost all of the information in databases is contained in tables. A table is
a simple 2D matrix of information. The rows in a table are called records and
the columns are called fields. The rows refer to individual entities. In a customer
table, the rows are customers. In an order table, the rows are orders. In a patient
table, the rows are patients. In a study table, the rows are the studies. The columns
are the attributes of an entity. In a patient database, the columns might be first
name, middle name, last name, date of birth, MRN, etc. In an administer dose
table, the columns might be radioisotope, radiopharmaceutical, administered
dose, time of injection, etc.
A key concept is that tables are very simple 2D matrixes of data. The
simplicity enables reliability. Almost all of the information in the database is in
a simple format that is easy to back up and easy to transfer between databases.
There is usually one field in each table that is unique for each row. That
unique field can be used for lookup. As that field allows access to the record, it
is called an accession number. The accession number for a patient table might
be the MRN; the accession number for a study table might be the study number;
443
CHAPTER 12
the accession number for a customer table would be the customer number. Some
databases use a combination of fields as the accession number, but the basic idea
is the same, there must be something unique which identifies each row.
12.6.1.2. Index
An index provides rapid access to the records (rows) of a table. The index
is separate from the table and is sorted to allow easy searching, often using a
tree structure. The tables themselves are usually not sorted; the records are just
added one after another. The index, not the table provides organization of the
records for fast access. Indexes are one of the important technologies provided by
a database vendor.
Indexes can be complicated and complex. However, there is no information
stored in an index. An index can be rebuilt from the tables. In fact, when
transferring data to a new database, the indexes are usually rebuilt. Indexes
are important for efficiency, but the only information they contain is how to
efficiently access the tables.
12.6.1.3. Relation
A key element of relational databases is relations. Relations connect one
table to another table. Conceptually, the relations form a very important part of a
database, and provide much of the complexity. However, the relations themselves
are actually very simple. A relation between the patient table and the study table
might be written:
patient.MRN=study.MRN
This says that the records in the patient table are linked to the records in the
study table by the MRN field.
A physician thinks of the database in terms of patients who have studies that
involve radiopharmaceuticals; however, a radiopharmacist thinks of isotopes and
radiopharmaceuticals. The relations facilitate these different points of view. The
physician can access the database in terms of patients and the radiopharmacist
can access the database in terms of radiopharmaceuticals. The same tables are
used; it is just the small amount of information contained in the relations that is
different.
444
12.6.2. Hospital information system

A hospital information system is a large distributed database. Data come
from clinical laboratories, nuclear medicine and financial systems, etc.
12.6.2.1. Admission, discharge, transfer
Most hospital information systems have an admission, discharge, transfer
(ADT) database. The ADT system is the master patient identification system.
Other systems use the ADT system for institution-wide, reliable, coordinated
identification of each patient.
12.6.2.2. Information gateway
Often, the goal in a business or a division of a larger organization is to
maintain a simple business model a small product line, etc. The simple
business model is often reflected in a simple database design. Medicine is
very complex and the principle of a simple, well structured design is a dream.
Generally, the different departments in the hospital have incompatible databases;
even within one department, there may be incompatible systems. One method
of ameliorating this problem is the development of Health Level Seven (HL7),
a standard message format for communicating between systems in a hospital.
However, the connections between systems still differ. If there are n systems,
communication between them becomes a problem that grows proportionally
to n2.
An information gateway ameliorates this problem. The only task of an
information gateway is to connect systems, translating messages so that they
can be understood by other systems. Each system only needs to connect to the
gateway, and the gateway communicates with all of the systems in the hospital.
The growth in complexity tends to increase more like n than n2.
12.6.3. Radiology information system
The radiology information system (RIS) supports scheduling, performing,
reporting procedure results and billing. When nuclear medicine is a division of
radiology, the RIS usually also functions as the nuclear medicine information
system. However, nuclear medicine procedures have some unique characteristics,
such as studies that extend over several days, which may not be well handled by
a general purpose RIS.
445
CHAPTER 12
12.6.4. Picture archiving and communication system

The image information from the imaging equipment is usually stored in a
picture archiving and communication system (PACS) that is separate from but
coordinated with the radiology/nuclear medicine information system. DICOM is
the predominant standard for PACSs.
12.6.4.1. Study
The top level of organization in DICOM is the study. This level of
organization comes from the organization of the health care system. Health
care providers request services from radiology/nuclear medicine by a request
for a consultation. Imaging or another service is performed and a report, ideally
including image information, is returned to the provider. Each study is linked to a
single patient, but a patient can have any number of separate studies.
12.6.4.2. Sequence
Sequence is the next level of organization in DICOM. Sequence comes
from a sequence of images; for a volume of image data, sequence is the top level
of organization. For example, it is common to collect a sequence of axial images.
A more general name for this level of organization would be dataset.
12.6.4.3. Image
Originally, DICOM used a 2D image as the basic atom. Other data
structures are composed of a number of images. A considerable amount of
metadata are included with each image, defining both the structure of the image,
patient information, data collection information and relation of the image to other
images. Somewhat more recently, a multiframe format was defined in DICOM.
One file may contain information from a volume of data, from a time series, from
a gated sequence, from different photopeaks, etc. Nuclear medicine tends to use
the multiframe format much more commonly than other modalities. Describing
the organization of a dataset in terms of multiple images is a particularly awkward
feature of DICOM.
12.6.4.4. N-dimensional data
It is unfortunate that DICOM selected the image as the basic atom of
organization (seeSection 12.5.4.3.3). Even before it was introduced, it was
apparent that an N-dimensional data model would be more appropriate. Nuclear
446
medicine and MRI often dealt with 1D curves, 2D images, 3D volumes,

4D gated or dynamic volumes, etc. However, radiology tended to be film based,
and volume data such as CT was anisotropic, so some of the early developers had
an image based orientation.
12.6.5. Scheduling
Scheduling is a much more complicated task than it may initially seem.
Several appointments in different departments may need to be scheduled at the
same time. Sequencing of studies may be important, e.g. thyroid imaging should
not be performed in proximity to iodinated contrast usage. Some studies require
a prior pregnancy test or other laboratory values. Prior data need to be made
available prior to performing some studies; for example, a prior electrocardiogram
should be available before a myocardial stress study.
Since scheduling needs to be coordinated between departments and
make use of hospital-wide information, it is logically a function of the hospital
information system. However, scheduling involves many issues that are local to
the modality, for example, availability of resources such as radiopharmaceuticals,
staff and equipment. Often, scheduling is local, making use of humans to provide
much of the hospital-wide coordination.
The scheduling system creates a worklist, a list of studies that need to be
performed. Most modern imaging equipment can use a worklist provided by the
RIS. When the technologist starts an imaging study, the appropriate patient is
picked from the worklist. Selecting a patient from a list results in far fewer errors
than re-entering all of the demographic identifier data for each study.
12.6.6. Broker
A broker is essentially an information gateway. The term information
gateway is generally used for a hospital-wide system. The term broker is
generally used when talking about a system within a radiology department. The
broker handles incompatible details related to RIS, PACS and imaging devices
that are local to radiology.
12.6.7. Security
Health information is private. Although the privacy of health information is
often a relativelyminor concern for the general public, it is a major concern for
politicians, who make rules about the security needed for medical information.
The theory of secure communication over insecure channels is thought to
be a solved problem. However, vigilance is necessary, because the practical
447
CHAPTER 12
implementations of the theory often have security holes that hackers can exploit.
Furthermore, with aggregation of private health information, the potential extent
of a security breach becomes catastrophic. Security depends not only on computer
systems, but also on humans who are often the weak link.
There needs to be a balance between the damage caused due to a security
breach and the expense of the security measures. Often, relatively low damage
situations are addressed with overly expensive systems, especially in terms
of lost productivity. The dominant effect of security should not be to prevent
authorized users from accessing information. Nuclear medicine tends to be a
relatively low risk environment, so in most circumstances the balance should
favour productivity over security.
BIBLIOGRAPHY
HUTTON, B.F., BARNDEN, L.R., FULTON, R.R., Nuclear medicine computers, Nuclear
Medicine in Clinical Diagnosis and Treatment (ELL, P.J., GAMBHIR, S.S., Eds), Churchill
Livingston, London (2004).
LEE, K.H., Computers in Nuclear Medicine: A Practical Approach, Society of Nuclear
Medicine, 2nd edn (2005).
PARKER, J.A., Image Reconstruction in Radiology, CRC Press, Boston, MA (1990).
PIANYKH, O.S., Digital Imaging and Communications in Medicine (DICOM): A Practical
Introduction and Survival Guide, Springer, Berlin (2008).
SALTZER, J.H., KAASHOEK, M.F., Principles of Computer System Design: An Introduction,
Morgan Kaufmann, Burlington (2009).
TODD-POKROPEK, A., CRADDUCK, T.D., DECONINCK, F., A file format for the
exchange of nuclear medicine image data: a specification of Interfile version 3.3, Nucl. Med.
Commun. 13 (1992) 673699.
448
CHAPTER 13
IMAGE RECONSTRUCTION
J. NUYTS
Department of Nuclear Medicine and Medical Imaging Research Center,
Katholieke Universiteit Leuven,
Leuven, Belgium
S. MATEJ
Medical Image Processing Group,
Department of Radiology,
University of Pennsylvania,
Philadelphia, Pennsylvania,
13.1. INTRODUCTION
This chapter discusses how 2D or 3D images of tracer distribution can
be reconstructed from a series of so-called projection images acquired with a
gamma camera or a positron emission tomography (PET) system [13.1]. This
is often called an inverse problem. The reconstruction is the inverse of the
acquisition. The reconstruction is called an inverse problem because making
software to compute the true tracer distribution from the acquired data turns out
to be more difficult than the forward direction, i.e. making software to simulate
the acquisition.
There are basically two approaches to image reconstruction: analytical
reconstruction and iterative reconstruction. The analytical approach is based
on mathematical inversion, yielding efficient, non-iterative reconstruction
algorithms. In the iterative approach, the reconstruction problem is reduced
to computing a finite number of image values from a finite number of
measurements. That simplification enables the use of iterative instead of
mathematical inversion. Iterative inversion tends to require more computer
power, but it can cope with more complex (and hopefully more accurate) models
of the acquisition process.
449
CHAPTER 13
13.2. ANALYTICAL RECONSTRUCTION

The (n-dimensional) radon transform maps an image of dimension n to
the set of all integrals over hyperplanes of dimension (n 1) [13.2]. Thus, in
two dimensions, the radon transform of image corresponds to all possible line
integrals of . In three dimensions, the radon transform contains all possible
plane integrals.
The (n-dimensional) X ray transform maps an image of dimension n to
the set of all possible line integrals. In all PET and in almost all single photon
emission computed tomography (SPECT) applications, the measured projections
can be well approximated as a subset of the (possibly attenuated) X ray transform,
because the mechanical (SPECT) or electronic (PET) collimation is designed to
acquire information along lines (the line of response (LOR), see Chapter11).
Consequently, reconstruction involves computing the unknown image from
(part of) its X ray transform. Figure13.1 shows PET projections, which are often
represented as a set of projections or a set of sinograms.
z
s
FIG.13.1. The relation between projections and sinograms in parallel-beam projection. The
parallel-beam (PET) acquisition is shown as a block with dimensions s, and z. A cross-section
at fixed yields a projection; a cross-section at fixed z yields a sinogram.
An important theorem for analytical reconstruction is the central slice (or

central section) theorem, which gives a relation between the Fourier transform of
an image and the Fourier transforms of its parallel projections. Below, the central
slice theorem for 2D is found as Eq.(13.7) and the 3D central section theorem
as Eq.(13.29).
450
The direct Fourier method is a straightforward application of the central

section theorem: it computes the Fourier transform of the projections, uses the
central section theorem to obtain the Fourier transform of the image and applies
the inverse Fourier transform to obtain the image. In practice, this method is
rarely used; the closely related filtered back projection (FBP) algorithm is far
more popular.
13.2.1. Two dimensional tomography
13.2.1.1. X ray transform: projection and back projection
In 2D, the radon transform and X ray transform are identical.
Mathematically, the 2D X ray (or radon) transform of the image can be written
as follows:
Y( s, ) =
=
( x, y) s=x cos + y sin dx dy
(13.1)
( s cos t sin , s sin + t cos ) dt
where the function is unity for the points on the LOR (s, ) and zero elsewhere.
It should be noted that with the notation used here, =0 corresponds to projection
along the y axis.
The radon transform describes the acquisition process in 2D PET and in
SPECT with parallel-hole collimation, if attenuation can be ignored. Assuming
that (x,y) represents the tracer distribution at transaxial slice Z through the
patient, then Y(s, ) represents the corresponding sinogram, and contains the z-th
row of the projections acquired at angles . Figure13.1 illustrates the relation
between the projection and the sinogram.
The X ray transform has an adjoint operation that appears in both analytical
and iterative reconstruction. This operator is usually called the back projection
operator, and can be written as:
B( x, y) = Backproj (Y( s, ))
=
=
Y( s, ) s=x cos + y sin ds (13.2)
Y( x cos + y sin , ) d
451
CHAPTER 13
The back projection is not the inverse of the projection, B(x,y) (x,y).
Intuitively, the back projection sends the measured activity back into the image
by distributing it uniformly along the projection lines. As illustrated in Fig.13.2,
projection followed by back projection produces a blurred version of the original
image. This blurring corresponds to the convolution of the original image with
the 2D convolution kernel 1 / x 2 + y 2 .
FIG.13.2. The image (left) is projected to produce a sinogram (centre), which in turn is back
projected, yielding a smoothed version of the original image.
13.2.1.2. Central slice theorem

The central slice theorem gives a very useful relation between the 2D
Fourier transform of the image and the 1D Fourier transform of its projections
(along the detector axis). Consider the projection along the y axis, =0, and its
1D Fourier transform:
Y( s,0) =
( F1Y)( s ,0) =
=
452
( s, t ) dt
Y( s,0)e i2 s s ds
(13.3)
( s, t )e
i2 s s
(13.4)
dt ds
and compare this to the 2D Fourier transform of the image (x,y):

( F2 )( x , y ) =
( x, y)e
i2 ( x x+ y y)
dx dy (13.5)
Both expressions are equal if we set y=0:

(F1Y)(s, 0)=(F2)(x, 0)
(13.6)
(F1Y)(s, 0) is the 1D Fourier transform of the projection along the y axis and
(F2)(x, 0) is a central slice along the x axis through the 2D Fourier transform
of the image. Equation (13.6) is the central slice theorem for the special case of
projection along the y axis. This result would still hold if the object had been
rotated or equivalently, the x and y axes. Consequently, it holds for any angle :
(F1Y)(s, )=(F2)(scos , ssin )
(13.7)
13.2.1.3. Two dimensional filtered back projection

The central slice theorem (Eq.(13.7)) can be directly applied to reconstruct
an unknown image (x,y) from its known projections Y(s, ). The 1D Fourier
transform of the projections provides all possible central slices through (F2)
(x, y) if Y(s, ) is known for all in an interval with a length of at least (Tuys
condition). Consequently, (F2)(x, y) can be constructed from the 1D Fourier
transform of Y(s, ). Inverse 2D Fourier transform then provides (x,y).
However, a basic Fourier method implementation with a simple
interpolation in Fourier space does not work well. In contrast, in the case of the
FBP algorithm derived below, a basic real-space implementation with a simple
convolution and a simple interpolation in the back projection works well. Inverse
Fourier transform of Eq.(13.5) yields:
( x, y) =
( F2 )( x , y )e
i2 ( x x+ y y)
d x d y (13.8)
This can be rewritten with polar coordinates as:

( x, y) =
( F2 )( cos , sin )e i2 ( x cos + y sin ) d (13.9)
453
CHAPTER 13
Application of the central slice theorem (Eq.(13.7)) and reversing the order
of integration finally results in:
( x, y) =
( F1Y)( , ) e i2 ( x cos + y sin ) d (13.10)
which is the FBP algorithm. This algorithm involves the following steps:
(a) Apply 1D Fourier transform to Y(s, ) to obtain (F1Y)(, );
(b) Filter (F1Y)(, ) with the so-called ramp filter ||;
(c) Apply the 1D inverse Fourier transform to obtain the ramp filtered
s, ) = ( F )( , ) e i2 s d ;
projections Y(
1
s, ) to obtain the
(d) Apply the back-projection operator Eq.(13.2) to Y(
desired image (x,y).
It should be noted that the ramp filter sets the DC component (i.e. the
amplitude of the zero frequency) of the image to zero, while the mean value of
the reconstructed image should definitely be positive. As a result, straightforward
discretization of FBP causes significant negative bias. The problem is reduced
with zero padding before computing the Fourier transform with fast Fourier
transform (FFT). Zero padding involves extending the sinogram rows with zeros
at both sides. This increases the sampling in the frequency domain and results
in a better discrete approximation of the ramp filter. However, a huge amount
of zero padding is required to effectively eliminate the bias completely. The
next paragraph shows how this need for zero padding can be easily avoided. It
should be noted that after inverse Fourier transform, the extended region may be
discarded, so the size of the filtered sinogram remains unchanged.
Instead of filtering in the Fourier domain, the ramp filtering can also be
implemented as a 1D convolution in the spatial domain. For this, the inverse
Fourier transform of || is required. This inverse transform actually does not exist,
but approximating it as the limit for 0 of the well behaved function ||e||
gives [13.3, 13.4]:
F 1( e ) =
2 (2 s) 2
(13.11)
( 2 + (2 s) 2 ) 2
1
(2 s) 2
for
s (13.12)
In practice, band limited functions are always worked with, implying that the
ramp filter has to be truncated at the frequencies =1/(2), where represents
454
the sampling distance. The corresponding convolution kernelh then equals

[13.3]:
h( s) = F 1( b( )) =
2
1 sin( s / )
1 sin( s / (2 ))
2 2 s /
4 2 s / (2 ) (13.13)
with b() = 1 if || 1/(2)

= 0 if || > 1/(2)
The kernel is normally only needed for samples s=n: h(n)=1/(42) if n=0,
h(n)=0 if n is even and h(n)=1/(n)2 if n is odd. The filter can either be
implemented as a convolution or the Fourier transform can be used to obtain
a digital version of the ramp filter. Interestingly, this way of computing the
ramp filter also reduces the negative bias mentioned above. The reason is that
this approach yields a non-zero value for the DC component [13.3]. When the
filtering is done in the frequency domain, some zero padding before FFT is still
recommended because of the circular convolution effects, but far less is needed
than with straightforward discretization of ||.
Although this is not obvious from the equations above, an algorithm
equivalent to FBP is obtained by first back projecting Y(s, ) and then applying a
2D ramp filter to the back projected image B(x,y) [13.4]:
B( x, y) =
Y( x cos + y sin , ) d (13.14)
( F2 )( x , y ) = x2 + y2 ( F2 B)( x , y ) (13.15)
This algorithm is often referred to as the back project-then-filter algorithm.

FBP assumes that the projections Y(s, ) are line integrals. As discussed in
Chapter11, PET and SPECT data are not line integrals because of attenuation,
detector non-uniformities, the contribution of scattered photons and/or random
coincidences, etc. It follows that one has to recover (good estimates of) the
line integrals by pre-correcting the data for these effects. However, a particular
problem is posed by the attenuation in SPECT because, different from PET,
the attenuation depends on the position along the projection line, precluding
455
CHAPTER 13
straightforward pre-correction. A detailed discussion of this problem is beyond

the scope of this contribution, but three solutions are briefly mentioned here:
(a) If it can be assumed that the attenuation is constant inside a convex body
contour, then FBP can be modified to correct for the attenuation. Algorithms
have been proposed by Bellini, Tretiak, Metz and others; an algorithm is
presented in Ref.[13.3].
(b) If the attenuation is not constant, an approximate correction algorithm
proposed by Chang can be applied [13.5]. It is a post-correction method,
applied to the image obtained without any attenuation correction.
To improve the approximation, the attenuated projection of the first
reconstruction can be computed, and the method can be applied again to the
difference of the measurement and this computed projection.
(c) Finally, a modified FBP algorithm, compensating for non-uniform
attenuation in SPECT, was found by Novikov in 2000. An equivalent
algorithm was derived by Natterer [13.6]. However, because this algorithm
was only found after the successful introduction of iterative reconstruction
in clinical practice, it has not received much attention in the nuclear
medicine community.
13.2.2. Frequencydistance relation
Several very interesting methods in image reconstruction, including Fourier
rebinning, are based on the so-called frequencydistance relation, proposed
by Edholm, Lewitt and Lindholm, and described in detail in Ref.[13.7]. This
is an approximate relation between the orthogonal distance to the detector and
the direction of the frequency in the sinogram. The relation can be intuitively
understood as follows.
Consider the PET acquisition of a point source, as illustrated in Fig 13.3.
Usually, the acquisition is described by rotating the projection lines while keeping
the object stationary. However, here the equivalent description is considered,
where projection is always along the y axis, and tomographic acquisition is
obtained by rotating the object around the origin. Suppose that the point is located
on the x axis when =0. When acquiring the parallel projections for angle ,
the point has polar coordinates (r, ), with r the distance from the centre of the
field of view (FOV) and the angle with the x axis. The distance to the x axis
is d=rsin . The corresponding sinogram Y(s, ) is zero everywhere, except on
the curve s=rcos (Fig.13.3). The complete sinogram is obtained by rotating
the point over 360 =. Consider a small portion of this curve, which can
be well approximated as a tangential line segment near a particular point (s, ),
456
d
x
FIG.13.3. The frequencydistance principle. Left: sinogram; right: vertical projection of a

point located at polar coordinates (r, ).
as illustrated in Fig.13.3. In the 2D Fourier transform of the sinogram, this line

segment contributes mostly frequencies in the direction orthogonal to the line
segment. This direction is represented by the angle , given by:
tan a =
(r cos ) = r sin = d . (13.16)
Thus, in the 2D Fourier transform (Fg) (s, ), the value at a particular

point (s, ) carries mostly information about points located at a distance
d=tan = /s from the line through the centre, parallel to the detector. This
relation can be exploited to apply distance dependent operations to the sinogram.
One example is distance dependent deconvolution, to compensate for the distance
dependent blurring in SPECT. Another example is Fourier rebinning, where data
from oblique sinograms are rebinned into direct sinograms.
13.2.3. Fully 3D tomography
13.2.3.1. Filtered back projection
Owing to the use of electronic collimation, the PET scanner can
simultaneously acquire information in a 4D space of line integrals. These are
457
CHAPTER 13
the so-called LORs, where each pair of detectors in coincidence defines a single
LOR. In this section, the discrete nature of the detection is ignored, since the
analytical approach is more easily described assuming continuous data. Consider
the X ray transform in 3D, which can be written as:
Y(u , s) =
( s + tu ) dt (13.17)
where the LOR is defined as the line parallel to and through the point s. The
vector is a unit vector, and the vector s is restricted to the plane orthogonal to ,
hence (, s) is 4D.
Most PET systems are either constructed as a cylindrical array of detectors
or as a rotating set of planar detector arrays and, therefore, have cylindrical
symmetry. For this reason, the inversion of Eq.(13.17) is studied for the case
where is restricted to the band 0 on the unit sphere, defined by u z sin 0 ,
as illustrated in Fig.13.4. It should be noted that only half of the sphere is actually
needed because Y(, s)=Y(, s), but working with the complete sphere is more
convenient.
FIG.13.4. Each point on the unit sphere corresponds to the direction of a parallel projection.
An ideal rotating gamma camera with a parallel-hole collimator only travels through the
points on the equator. An idealized 3D PET system would also acquire projections along
oblique lines; it collects projections for all points of the set . The set , defined by 0, is
the non-shaded portion of the unit sphere. To recover a particular frequency (of the Fourier
transform of the object), at least one point on the circle C is required.
With 0=0, the problem reduces to 2D parallel projection (for multiple

slices), which was shown to have a unique solution. It follows that with
458
|0| > 0, the problem becomes overdetermined, and there are infinitely many ways
to compute the solution. This can be seen as follows. Each point of corresponds
to a parallel projection. According to the central slice theorem, this provides a
central plane perpendicular to of the 3D Fourier transform L() of (x). Thus,
the set 0 (i.e. all points on the equator of the unit sphere in Fig.13.4) provides
all planes intersecting the z axis, which is sufficient to recover the entire image
(x) via inverse Fourier transform. The set 0 with 0 > 0 provides additional
(oblique) planes through L(), which are obviously redundant. A simple solution
would be to select a sufficient subset from the data. However, if the data are
noisy, a more stable solution is obtained by using all of the measurements. This
is achieved by computing L() from a linear combination of all available planes:
L( ) =
Y (u , )H(u , )(u , ) du (13.18)

0
Here, Y (u , ) is the 2D Fourier transform with respect to s of the projection

Y(u , s) . The Dirac function (u , ) selects the parallel projections which are
perpendicular to (i.e. the points on the circle C in Fig.13.4). Finally, the filter
H(u , ) assigns a particular weight to each of the available datasets Y (u , ) . The
combined weight for each frequency should equal unity, leading to the filter
equation:
H(u , )(u , ) du = 1 (13.19)

0
A solution equivalent to that of unweighted least squares (LS) is obtained by

assigning the same weight to all available data [13.8]. This results in the Colsher
filter which can be written as:
H C (u , ) = / (2 )
if sin sin 0
= / (4arcsin(sin 0 / sin ) if sin > sin 0
(13.20)
where is the angle between and the z axis: z/||=cos . The direct Fourier
reconstruction method can be applied here, by straightforward inverse Fourier
transform of Eq.(13.18). However, an FBP approach is usually preferred, which
can be written as:
( x ) =
du Y F (u , x ( x u )u ) (13.21)
0
459
CHAPTER 13
Here, YF is obtained by filtering Y with the Colsher filter (or another filter
satisfying Eq.(13.19): Y F (u , s) = F 1(H C (u , )Y (u , )). The coordinate
s = x ( x u )u is the projection of the point x on the plane perpendicular to ; it
selects the LOR through x in the parallel projection .
13.2.3.2. The reprojection algorithm
The previous analysis assumed that the acceptance angle 0 was a constant,
independent of x. As illustrated in Fig.13.5, this is not the case in practice. The
acceptance angle is maximum for the centre of the FOV, becomes smaller with
increasing distance to the centre and vanishes near the axial edges of the FOV. In
other words, the projections are complete for orthogonal to the z axis (these are
the 2D multislice parallel-beam projections) and are truncated for the oblique
parallel projections. The truncation becomes more severe for more oblique
projections (Fig.13.5).
(a)
(b)
(c)
FIG.13.5. An axial cross-section through a cylindrical PET system, illustrating that the
acceptance angle is position dependent (a). Oblique projections are truncated (b). In the
reprojection algorithm, the missing oblique projections (dashed lines) are computed from a
temporary multislice 2D reconstruction (c).
As the acceptance angle is position dependent, the required filtering

is position dependent as well, and cannot be implemented as a shift-invariant
convolution (or Fourier filter). Several strategies for dealing with this truncation
have been developed. One approach is to subdivide the image into a set of
regions, and then optimize a shift-invariant filter in each of the regions. The filter
is determined by the smallest acceptance angle of the region, so some of the data
460
will not be used. A good compromise betweenminimum data loss and practical
implementation must be sought [13.9].
Another approach is to start with a first reconstruction, using the smallest
acceptable angle over all positions x in the FOV. This usually means that only
the parallel projections orthogonal to the z axis are used. The missing oblique
projection values are computed from this first reconstruction (Fig.13.4) and used
to complete the measured oblique projections. This eliminates the truncation, and
the 3D FBP method of the previous section can be applied. This method [13.10]
was the standard 3D PET reconstruction method for several years, until it was
replaced by the faster Fourier rebinning approach (seebelow).
13.2.3.3. Rebinning techniques
The complexity (estimated as the number of LORs) increases linearly
with the axial extent for 2D PET, but quadratically for 3D PET. To keep the
processing time acceptable, researchers have sought ways to reduce the size of
the data as much as possible, whileminimizing the loss of information induced
by this reduction.
Most PET systems have a cylindrical detector surface: the detectors are
located on rings with radius R, and the rings are combined in a cylinder along the
z axis. The data are usually organized in sinograms which can be written as:
YP ( s, , z, z ) =
dt ( s cos + tu x , s sin + tu y , z + tu z ) (13.22)
where is a unit vector in the direction of the LOR:

u = u / u
with
u = ( sin ,cos , z / (2 R 2 s 2 ))
The parameter s is the distance between the LOR and the z axis. The LOR
corresponds to a coincidence between detector points with axial positions z z/2
and z + z/2. Finally, is the angle between the y axis and the projection of the
LOR on the xy plane. The coordinates (s, , z) are identical to those often used in
2D tomography. It should be noted that, in practice, s R and, as a result, the
direction of the LOR, the vector , is virtually independent of s. In other words,
a set of LORs with fixed z can then be treated as a parallel projection with good
approximation. LORs with z =0 are often called direct LORs, while LORs
with z 0 are called oblique.
The basic idea of rebinning algorithms is to compute estimates of the
direct sinograms from the oblique sinograms. If the rebinning algorithm is good,
most of the information from the oblique sinograms will go into these estimates.
461
CHAPTER 13
As a result, the data have been reduced from a complex 3D geometry into a
much simpler 2D geometry without discarding measured signal. The final
reconstruction can then be done with 2D algorithms, which tend to be much
faster than fully 3D algorithms. A popular approach is to use Fourier rebinning,
followed by maximum-likelihood reconstruction.
13.2.3.4. Single slice and multislice rebinning
The simplest way to rebin the data is to treat oblique LORs as direct LORs
[13.11]. This corresponds to the approximation:
YP ( s, , z, z ) YP ( s, , z,0) (13.23)
The approximation is only exact if the object consists of points located on

the z axis, and it introduces mis-positioning errors that increase with increasing
distance to the z axis and increasing z. Consequently, single slice rebinning is
applicable when the object is small and positioned centrally in the scanner or
when z is small. The axial extent of most current PET systems is too large to
rebin an entire 3D dataset with Eq.(13.23). However, single slice rebinning is
used on all PET systems to reduce the sampling of the z dimension in the 3D
data, by combining sinograms with similar z. This typically reduces the data
size by a factor of about ten, when compared to the finest possible sampling.
Application of Eq.(13.23) obviously causes blurring in the z direction,
to a degree proportional to the distance from the z axis. However, it may also
cause severe inconsistencies in the sinograms, producing blurring artefacts
in the xy planes of the reconstructed images as well. Lewitt et al. [13.12]
proposed distributing the oblique LOR values YP (s, , z, z ) over all LORs
with z [z zR f / (2 R), z + zR f / (2 R)] , i.e. over all slices intersected by the
LOR, and within an FOV with radius Rf. This so-called multislice rebinning
reduces the inconsistencies in the sinograms, eliminating most of the xy blurring
artefacts in the reconstruction. Unfortunately, the improvement comes at the cost
of strong axial blurring. This blurring depends strongly on z, but it is found to be
approximately independent of x and y. A z-dependent 1D axial filter is applied to
reduce this axial blurring [13.12]. Multislice rebinning is superior to single slice
rebinning, but the noise characteristics are not optimal.
462
13.2.3.5. Fourier rebinning

Fourier rebinning [13.13] is based on the frequencydistance principle,
which was explained previously. The Fourier rebinning method is most simply
formulated when the projection is written as follows:
Y( s, , z, ) =
dt ( s cos t sin , s sin + t cos , z + t ) (13.24)
where
=tan ;
is the angle between the LOR and the xy plane;
and the integration variable t is the distance between the position on the LOR and
the z axis.
It follows that:
Y(s, , z, ) =
YP ( s, , z, z = 2 R 2 s 2 )
1+ 2
YP (s, , z, z = 2 R)
1+ 2
(13.25)
(13.26)
where the approximation is valid whenever s R. In this case, no interpolation

is needed; it is sufficient to scale the PET data YP with the weight factor 1 + 2 .
Fourier rebinning uses the frequencydistance principle to find the
distance d corresponding to a particular portion of the oblique sinogram. As
illustrated in Fig.13.6, distance is used to locate the direct sinogram to which this
portion should be assigned. Denoting the 2D Fourier transform of Y with respect
to s and as Y, this can be written as:
Y ( s , , z, ) Y ( s , , z
,0) (13.27)
This equation explains how to distribute frequency components from a particular

oblique sinogram into different direct sinograms. Frequencies located on the line
463
CHAPTER 13
FIG.13.6. Fourier rebinning: the distance from the rotation axis is obtained via the frequency
distance principle. This distance is used to identify the appropriate direct sinogram.
= s in the oblique sinogram z can be assigned to that same line in the direct
sinogram z+d.
The final rebinning algorithm (often called FORE) is obtained by

averaging all of the available estimates of the direct sinogram:
Y ( s , , z,0)
max
s 0
Y (0,0, z,0)
if
s 0, 0 (13.28)
if
/ s > R f
d Y ( s , , z +
if
max
, )
It should be noted that the rebinning expression is only valid for large s.
In the low frequency range, only the direct sinogram is used. The last line of
Eq.(13.28) holds because the image (x, y, z) is assumed to be zero outside the
FOV x 2 + y 2 > R f .
A more rigorous mathematical derivation of the frequencydistance
relation is given in Ref.[13.14]. Alternative derivations based on exact rebinning
expressions are given in Ref.[13.13].
464
After Fourier rebinning, the resulting 2D dataset can be reconstructed

with any 2D reconstruction algorithm. A popular method is the combination of
Fourier rebinning with a 2D statistical reconstruction algorithm.
13.2.3.6. Exact rebinning methods
Fourier rebinning is an approximate method, but was found to be sufficiently
accurate for apertures up to 0=25, and it is, therefore, largely sufficient for
most current PET systems. However, there is a tendency towards still larger
acceptance angles, and a more exact Fourier rebinning algorithm may be needed
in the future. An example of an exact rebinning algorithm is FOREX [13.13]. It
is exact in the sense that the rebinning expression is exact for the continuous 3D
X ray transform.
According to the central section theorem, the 2D Fourier transform of a
projection Y(s, , z, ) equals a cross-section through the 3D Fourier transform
of the image (x, y, z):
Y13 ( s , , z , ) = L( s cos + z sin , s sin z cos , z )
(13.29)
The subscript of Y13 denotes a Fourier transform with respect to s and z. Defining:
= arctan( z / s )
' s = s 1 + 2v z2 s2
Equation (13.29) can be rewritten as:

Y13 ( s , , z , ) = L( ' s cos( ), ' s sin( ), z ) (13.30)
Taking the 1D Fourier transform with respect to yields:

Y123 ( s , , z , ) = e
2
0
L( ' s cos , ' s sin , z ) d
(13.31)
By comparing the expressions for Y123 ( s , , z , ) and Y123 ( s , , z ,0), one

finally obtains:
Y123 ( s , , z , ) = e
Y123 ( ' s , , z ,0) (13.32)
A problem of FOREX is that it needs the 1D Fourier transform along z, which

cannot be computed for truncated projections. Similar as with 3D filtered back
465
CHAPTER 13
projection, the problem can be avoided by completing the truncated projections

with synthetic data. Fortunately, Eq.(13.32) can be used in both ways, and allows
estimation of (missing) oblique sinograms from the available direct sinograms.
The resulting algorithm is slower than FORE, but still considerably faster than
3D FBP with reprojection.
13.2.4. Time of flight PET
In time of flight (TOF) PET, the difference in arrival time of the two
detected photons is used to estimate the position of their emission along the
LOR. The uncertainty in the time estimation results in a similar uncertainty in the
position estimation, which can be well modelled as a Gaussian distribution. As
a result, the TOF projections correspond to Gaussian convolutions along lines,
rather than to line integrals, as illustrated in Fig.13.7.
FIG.13.7. Time of flight projection can be well modelled as a 1D Gaussian convolution in

the direction of the line of response.
The corresponding TOF back projection corresponds to convolving the

measured data with the same 1D Gaussians, followed by summation over all
angles.
Recall from Eq.(13.2) that the regular projection followed by the regular
back projection corresponds to a convolution with a blurring filter:
BnonTof ( x, y) =
466
1
2
x + y2
(13.33)
The Fourier transform of 1/ x 2 + y 2 equals 1 / x2 + y2 . Consequently, this

blurring can be undone by the ramp filter x2 + y2 , which can be applied either
before or after back projection (seeSection 13.2.1).
If TOF is the standard deviation of the TOF-blurring kernel, then TOF
projection followed by TOF back projection corresponds to convolution with the
blurring kernel:
BTof ( x, y) =
Gauss ( x, y, 2 TOF )
x 2 + y2
(13.34)
x 2 + y 2
1
(13.35)
exp
2
4 TOF
x 2 + y 2 2 TOF
1
It should be noted that the Gaussian in the equation above has a standard deviation
of 2 TOF . This is because the Gaussian blurring is present in the projection
and in the back-projection. The filter required in TOF PET FBP is derived by
inverting the Fourier transform of BTOF, and equals:
TOF_recon_filter ( ) =
1

2
2
exp(2 2 TOF
2 )I 0 (2 2 TOF
2)
(13.36)
where I 0 is the zero order modified Bessel function of the first kind.
This FBP expression is obtained by using the natural TOF back projection,
defined as the adjoint of the TOF projection. This back projection also appears in
LS approaches, and it has been shown that with this back projection definition,
FBP is optimal in an (unweighted) LS sense [13.15]. However, TOF PET data
are redundant and different back projection definitions could be used; they would
yield different expressions for BTOF(x,y) in Eq.(13.34) and, therefore, different
TOF reconstruction filters.
Just as for non-TOF PET, exact and approximate rebinning algorithms for
TOF PET have been derived to reduce the data size. As the TOF information
limits the back projection to a small region, the errors from approximate rebinning
are typically much smaller than in the non-TOF case.
467
CHAPTER 13
13.3. ITERATIVE RECONSTRUCTION

13.3.1. Introduction
13.3.1.1. Discretization
In analytical reconstruction, it is initially assumed that the unknown object

can be represented as a function ( x) with x R 3 , and that the acquired data
can be represented as a function Y(s, ) with s R 2 and a unit vector in R 2
or R 3 . The reconstruction algorithm is then derived by mathematical inversion
(assuming some convenient properties for and Y), and finally the resulting
algorithm is discretized to make it ready for software implementation. In iterative
reconstruction, one usually starts by discretizing the problem. This reduces the
reconstruction problem to finding a finite set of unknown values from a finite
set of equations, a problem which can be solved with numerical inversion.
The advantage of numerical inversion is that only a model for the acquisition
process is needed, not for its inverse. That makes it easier (although it may still
be non-trivial) to take into account some of the undesired but unavoidable effects
that complicate the acquisition, such as photon attenuation, position dependent
resolution, gaps between the detectors and patient motion.
After discretization, the unknown image values and the known measured
values can be represented as column vectors and y. The PET or SPECT
acquisition process is characterized by the system matrix A and an additive
contribution b , and n is the measurement noise:
y = A + b + n
or
yi =
A
ij
+ bi + n i ,
i = 1,..., I
(13.37)
j=1
The symbol yi denotes the number of photons measured at LOR i, where the
index i runs over all of the sinogram elements (merging the three or four sinogram
dimensions into a single index). The index j runs over all of the image voxels,
and Aij is the probability that a unit of radioactivity in j gives rise to the detection
of a photon (SPECT) or photon pair (PET) in LOR i. The estimate of the additive
contribution is denoted as b . This estimate is assumed to be noise-free and
includes, for example, scatter and randoms in PET or cross-talk between different
energy windows in multitracer SPECT studies. Finally, ni represents the noise
contribution in LOR i.
Image reconstruction now consists of finding , given A, y and b , and a
statistical model for n.
468
For further reading about this subject, the recent review paper on iterative
reconstruction by Qi and Leahy [13.16] is an ideal starting point.
13.3.1.2. Objective functions
The presence of the noise precludes exact reconstruction. For this reason,
the reconstruction is often treated as an optimization task: it is assumed that a
useful clinical image can be obtained by maximizing a well chosen objective
function. When the statistics of the noise are known, a Bayesian approach can be
applied, searching for the image l that maximizes the conditional probability on
the data:
= arg max p( | y)
= arg max
p( y | ) p( )
(13.38)
p( y)
= arg max p( y | ) p( )
= arg max(ln p( y | ) + ln p( )) (13.39)
The second equation is Bayes rule. The third equation holds because y does not
depend on , and the fourth equation is valid because computing the logarithm
does not change the position of the maximum. The probability p(y|) gives the
likelihood of measuring a particular sinogram y, when the tracer distribution
equals . This distribution is often simply called the likelihood. The probability
p() represents the a priori knowledge about the tracer distribution, available
before PET or SPECT acquisition. This probability is often called the prior
distribution. The knowledge available after the measurements equals p(y|)p()
and is called the posterior distribution. To keep things simple, it is often assumed
that no prior information is available, i.e. p(|y) p(y|). Finding the solution
then reduces to maximizing the likelihood p(y|) (or its logarithm). In this
section, maximum-likelihood algorithms are discussed. Maximum a posteriori
(MAP) algorithms are discussed in Section 13.3.5, as a strategy to suppress noise
propagation.
A popular approach to solve equations of the form of Eq.(13.37) is LS
estimation. This is equivalent to a maximum-likelihood approach, if it is assumed
that the noise is Gaussian with a zero mean and a fixed, position independent
469
CHAPTER 13
standard deviation . The probability to measure the noisy value yi when the
Aij l j + bi then equals:
expected value was
pLS ( y i |
( y (
1
Aij l j + bi ) =
exp
2
A l
ij
2 2
+ bi )) 2
(13.40)
As the noise in the sinogram is not correlated, the likelihood (i.e. the probability
of measuring the entire noisy sinogram y) equals:
pLS ( y | ) = pLS ( y | A + b ) =
p
i
LS ( y i
A
ij
+ bi ) (13.41)
It is more convenient to maximize the logarithm of pLS; dropping constants, the

objective function LLS is finally obtained:
( y ( A
LLS =
ij
+ bi )) 2 = ( y ( A + b ))'( y ( A + b ))
(13.42)
where the prime denotes matrix transpose. Setting the first derivatives with
respect to j to zero for all j gives:
A'( y A b ) = 0
( A' A)1 A'( y b )
(13.43)
is the discrete projection; its

provided that A' A is non-singular. The operator A A
transpose A' is the discrete back projection. Its analytical counterpart was given
in Eq.(13.2) and illustrated in Fig.13.1. The same figure shows that the operator
A' A behaves as a blurring filter.
A A
FIG.13.8. The image of point sources is projected and back projected again along ideal
parallel beams. This yields a shift-invariant blurring.
470
Figure13.8 is similar, but illustrates A and A' A on an image of three point

sources, using ideal parallel-beam projection. The figure shows the resulting point
spread functions of A' A for each of the point sources. They are identical: for ideal
parallel-beam projection, A' A is shift-invariant, equivalent to a convolution. It
follows that ( A' A)1 is the corresponding shift-invariant deconvolution, which
is easily computed via the Fourier transform. In this situation, LS reconstruction
(Eq.(13.43)) is the discrete equivalent of the back project-then-filter algorithm
(Eq.(13.15)), applied to the data after pre-correction for b .
Figure13.9 illustrates A and A' A for a projector that models the position
dependent blurring of a typical parallel-beam SPECT collimator. The blurring
induced by A' A is now shift-variant it cannot be modelled as a convolution
and its inverse cannot be computed with the Fourier transform. For real life
problems, direct inversion of A' A is not feasible. Instead, iterative optimization
is applied to find the maximum of Eq.(13.42).
A A
FIG.13.9. The image of point sources is projected and back projected again with collimator
blurring. This yields a shift-variant blurring.
It is known that the number of detected photons is subject to Poisson noise,

not to uniform Gaussian noise. The Poisson distribution can be well approximated
with a Gaussian distribution, where the variance of the Gaussian equals its mean.
With this approximation, must be replaced by i in Eq.(13.40) because now
there is a different Gaussian distribution for every sinogram pixel i. Proceeding
as before, the weighted least squares (WLS) objective function is:
LWLS =
( yi (
A + b ))
ij j
i2
(13.44)
= ( y ( A + b ))'C y 1( y ( A + b ))
where Cy is the covariance matrix of the data.
471
CHAPTER 13
For emission tomography, it is a diagonal matrix (all covariances are zero)

with elements Cy[i, i]=i2. The corresponding WLS reconstruction can be written
as:
= ( A'C y 1 A) 1 A'C y 1( y b ) (13.45)
The operator A'C y 1 A is always shift-variant, even for ideal parallel-beam

tomography. This is illustrated in Fig.13.10. The noise-free sinogram y is
computed for a particular activity distribution. Setting Cy=diag ( y ), the operator
A'C y 1 A can be analysed by applying it to the image of a few point sources,
called x in the figure. The image x is projected, the sinogram Ax is divided by
y on a pixel basis and the result is back projected. Clearly, position dependent
blurring is obtained. Consequently, iterative optimization must be used for WLS
reconstruction.
Emission image
Ax
Cy1Ax
A'Cy1Ax
FIG.13.10. The operator A'C y A is derived for a particular activity distribution (top left)
and then applied to a few point sources x. Although ideal parallel-beam projection was used,
shift-variant blurring is obtained.
1
In practice, because there is only a noisy sinogram y, the noise-free

sinogram y must be estimated to find Cy. There are basically two approaches.
In the first approach, y is estimated from y, e.g. by smoothing y to suppress
the noise. In the second approach, y is estimated as A (k ) + b during the
iterative optimization, where (k) is the estimate of the reconstruction available at
iteration k. A drawback of the first approach is that the noise on the data affects
the weights, with a tendency to give higher weight when the noise contribution
472
happens to be negative. A complication of the second approach is that it makes

i a function of . In this case, the normalizing amplitude 1/ ( 2 i ) of the
Gaussians cannot be dropped, implying that an additional term ln i should
i
be added to Eq.(13.44).
It is possible to use the Poisson distribution itself, instead of approximating
it with Gaussians. The probability of the noise realization yi then becomes:
(
e
pML ( y i |
A
ij
Aij j +bi )
j
A
ij
+ bi ) y i
+ bi ) =
(13.46)
yi !
Proceeding as before, the log-likelihood function is:
ln
pLS ( y i |
LML =
Aij j + bi ) =
y ln( A
i
ij j
+ bi ) (
y ln( A
i
ij
ij j
+ bi ) (
A
ij
+ bi ) ln y i !
+ bi ) (13.47)
It should be noted that the term ln yi! can be dropped, because it is not a function
of . As LML is a non-linear function of , the solution cannot be written as a
product of matrixes. However, it is sometimes helpful to know that the features
of the Poisson-objective function are often very similar to those of the WLS
function (Eq.(13.44)).
13.3.2. Optimization algorithms
Many iterative reconstruction algorithms have been proposed to optimize
the objective functions LWLS and LML. Here, only two approaches are briefly
described: preconditioned conjugate gradient methods and optimization transfer,
with expectation maximization (EM) as a special case of the latter.
13.3.2.1. Preconditioned gradient methods
The objective function will be optimized when its first derivatives are zero:
y i =
ij j
+ bi (13.48)
473
CHAPTER 13
LWLS ( )
=
j
LML ( )
=
j
ij
y i y i
(13.49)
i2
y i y i
(13.50)
y i
ij
The optimization can be carried out by a steepest ascent method, which can be
formulated as follows:
d k = L( k1 )
(13.51)
k = k1 + k d k
k = arg max L( k1 + d k )
where the superscripts k and k1 denote the iteration numbers and L

L is the
vector of the first derivatives of L with respect to j.
Steepest gradient ascent is known to be suboptimal, requiring many
iterations for reasonable convergence. To find a better update, it is required that
after the update, the first derivatives of L are zero as intended. Approximating
this with a first order Taylor expansion yields:
L( k1 + p k ) = 0
L( k1 ) + Hp k 0
(13.52)
p k H 1L( k1 ) = H 1d k
where the Hessian H is the matrix of second derivatives of L. This is obviously a

very large matrix, but its elements are relatively easy to compute:
for WLS: H jk =
for ML: H jk =
Aij Aik
i2
Aij Aik y i
y i2
= ( A'C y 1 A)[ j, k ] (13.53)
y i
( A'C y 1 A)[ j, k ]
474
Aij Aik
if
(13.54)
y y (13.55)
For a Gaussian likelihood, Eq.(13.52) is in fact exact, and a single iteration

would suffice. As shown before, however, it is usually impossible to compute
H1. Instead, approximations to the Hessian (or other heuristics) can be used to
obtain a good M to derive a so-called preconditioned gradient ascent algorithm:
d k = L( k1 )
(13.56)
=
k
k1
+ k Md
To ensure that the convergence is preserved, the matrix M must be symmetric

positive definite (it should be noted that H1 is symmetric positive definite,
since H is symmetric negative definite, if A has maximum rank).
A simple way to obtain a reasonable M is to use only the diagonal elements
of H: Mii=1/Hii and Mij=0 if i j. A more sophisticated approach is discussed
in Ref.[13.17]: a circulant, i.e. shift-invariant approximation of the Hessian is
proposed. Such an approximation is easily computed by fixing j at a particular
location in the image in Eqs(13.53) or (13.54), which yields an image that
can be considered as the point spread function of a convolution operator. This
shift-invariant operator is then inverted via the Fourier transform, yielding
a non-diagonal matrix M. For cases where the true Hessian depends heavily
on position, the approach could be repeated for a few well chosen positions j,
applying linear interpolation for all other positions.
13.3.2.2. Conjugate gradient methods
Figure13.11 shows the convergence of the steepest gradient ascent algorithm
for a nearly quadratic function of two variables. In every iteration, the algorithm
starts moving in the direction of the maximum gradient (i.e. perpendicular to the
isocontour), and keeps moving along the same line until a maximum is reached
(i.e. until the line is a tangent to the isocontour). This often leads to a zigzag line,
requiring many iterations for good convergence.
The conjugate gradient algorithm is designed to avoid these oscillations
[13.18]. The first iteration is identical to that of the steepest gradient ascent.
However, in the following iterations, the algorithm attempts to move in a direction
for which the gradient along the previous direction(s) remains the same (i.e. equal
to zero). The idea is to eliminate the need for a new optimization along these
previous directions. Let dold be the previous direction and H the Hessian matrix
(i.e. the second derivatives). It is now required that the new direction dnew be such
that the gradient along dold does not change. When moving in direction dnew, the
475
CHAPTER 13
FIG.13.11.
dotted
lines
isocontours
of objective
the objective
function.
The line
solidshows
line shows
FIG.
13.11. TheThe
dotted
lines
are are
isocontours
of the
function.
The solid
convergence
of the
steepest
gradient
ascent
algorithm,
the dashed
line
the convergence
of
thethe
convergence
of the
steepest
gradient
ascent
algorithm,
the dashed
line the
convergence
of
conjugent
gradient
ascent.
It should
noted
starting
points
equivalent
because of
conjugent
gradient
ascent.
It should
be be
noted
thatthat
thethe
starting
points
are are
equivalent
because
ofthe
the symmetry.
symmetry. The objective
p=
2.15.
, with
p=2.15.
objectivefunction
functionequals
equals(a(a| x| x x0x|0p | pb+b| y| y y0y|0p )| p,)with
gradient
will change
(using aisquadratic
as Hd
new. Requiring
The conjugate
gradient algorithm
designed toapproximation)
avoid these oscillations
[13.18].
The first that
the resulting
change
along
doldsteepest
is zero gradient
yields the
condition:
iteration
is identical
to that
of the
ascent.
However, in the following
iterations, the algorithm attempts to move in a direction for which the gradient along the
previous direction(s) remains
the same (i.e. equal to zero). The idea is to eliminate the need
(13.57)
d ' Hd new = 0
for aold
new optimization along these previous directions. Let dold be the previous direction and
H the Hessian matrix (i.e. the second derivatives). It is now required that the new direction
dnew be such
the gradient
dold doesbynotthe
change.
When
moving
in directionin
dnew
, the
Thisthat
behaviour
isalong
illustrated
dashed
line
in Fig.13.11:
the
second
gradient will change (using a quadratic approximation) a Hdnew. Requiring that the resulting
iteration,
the
algorithm
moves
in
a
direction
such
that
the
trajectory
cuts
the
change along dold is zero yields the condition:
isocontours at the same angle as in the starting point. For a quadratic function
(13.57)As the
H d new = 0. convergence is obtained after no more than n iterations.
in dn' olddimensions,
function in Fig.13.11 is not quadratic, more than two iterations are required for
This behaviour is illustrated by the dashed line in Fig. 13.11: in the second iteration, the
full convergence.
algorithm moves in a direction such that the trajectory cuts the isocontours at the same angle
newpoint.
direction
can be easily
computed
from the
previousisones,
without
as in theThe
starting
For a quadratic
function
in n dimensions,
convergence
obtained
after
no more thanofn the
iterations.
As the
in Fig. 13.11 is algorithm
not quadratic,
Hessian
H .function
The PolakRibiere
is more
giventhan
bytwo
[13.18]:
computation
iterations are required for full convergence.
The new direction can be easily computed from the previous ones, without computation of
new =
old )PolakRibiere algorithm is given by [13.18]:
thegHessian
HL(. The
( g new g old )' g new

g 'old g old
d new = g new + d old
= arg max L( old + d new )
new = old + d new
476
(13.58)
This algorithm requires storage of the previous gradient gold and the
previous search direction dold. In each iteration, it computes the new gradient and
search direction, and applies a line search along the new direction.
13.3.2.3. Preconditioned conjugate gradient methods
Both techniques mentioned above can be combined to obtain a fast
reconstruction algorithm, as described in Ref.[13.17]. The preconditioned
conjugate gradient ascent algorithm (with preconditioning matrix M) can be
written as follows:
g new = L( old )
p new = Mg new
( g new g old )' p new

g 'old p old
(13.59)
d new = p new + d old
= arg max L( old + d new )
new = old + d new
13.3.2.4. Optimization transfer

The log-likelihood function (Eq.(13.47)) can be maximized by setting its
gradients (Eq.(13.50)) to zero for all j=1J. A problem is that each of these
derivatives is a function of many voxels of , which makes the set of equations
very hard to solve. The idea of optimization transfer is to replace the problematic
log-likelihood function with another function () that leads to a simpler set of
equations, usually one where the derivative with respect to j is only a function
of j and not of the other voxels of . That makes the problem separable into
J 1D optimizations, which are easily solved. Ideally, and L should have the
same optimum, but that is asking for too much. The key is to design () in
such a way that maximization of () is guaranteed to increase L(). This leads
to an iterative algorithm, since new functions will have to be designed and
477
CHAPTER 13
maximized repeatedly to maximize L. At iteration k, the surrogate function ()

needs to satisfy the following conditions (illustrated in Fig.13.12):
( (k ) ) = L( (k ) ) (13.60)
( X ) L( X ) (13.61)
Likelihood
Current
New
FIG.13.12. Optimization transfer: a surrogate function is designed, which is equal to the

likelihood in the current reconstruction, and less or equal everywhere else.
It follows that the new reconstruction image (k+1) which maximizes ()

has a higher likelihood than (k):
L( (k ) ) = ( (k ) ) ( (k +1) ) L( (k +1) ) (13.62)
Several algorithms for maximum-likelihood and MAP reconstruction in emission

and transmission tomography have been developed with this approach. De
Pierro [13.19] has shown how the well known maximum-likelihood expectationmaximization (MLEM) algorithm can be derived using the optimization transfer
principle. He also showed how this alternative derivation provides a natural way
to extend it to an MAP algorithm.
478
13.3.3. Maximum-likelihood expectation-maximization

13.3.3.1. Reconstruction from sinogram data
There are many ways to derive the MLEM algorithm, including the original
statistical derivation by Shepp and Vardi [13.20] (based on the work by Dempster
et al. [13.21]) and the optimization transfer approach by De Pierro [13.19]. Only
the EM recipe is given below.
Recall that we wish to find the image that maximizes the likelihood
function LML of Eq.(13.47). The EM does this in a remarkable way. Instead of
concentrating on LML, an alternative (different) likelihood function is derived by
introducing a set of so-called complete data xij, defined as the number of photons
that were emitted at voxel j and detected in LOR i during the measurement.
These unobserved data are complete in the sense that they describe in more
detail than the observed data yi what happened during the measurement. These
variables xij are Poisson distributed. Just as for the actual data yi, one can write the
log-likelihood function for observing the data xij while x ij = Aij j were expected:
L x ( ) =
x
i
ij
ln( Aij j ) Aij j (13.63)
However, this likelihood cannot be computed, because the data xij are not
available. The emission measurement only produces sums of the complete data,
since:
yi =
A x
ij ij
+ bi (13.64)
where bi represents the actual (also unobserved) additive contribution bi in LOR i.

The EM recipe prescribes computing the expectation of Lx, based on the
available data and on the current reconstruction (k). Based on the reconstruction
alone, one would write E( x ij | (k ) ) = Aij j(k ) . However, it is also known that
xij should satisfy Eq.(13.64). It can be shown that this leads to the following
estimate:
E( x ij | (k ) , y) =
yi
Aij j(k ) + bi
Aij j(k ) (13.65)
where bi is the noise-free estimate of bi, which is assumed to be available.
479
CHAPTER 13
Inserting this in Eq.(13.63) produces the expectation of Lx() and completes

the expectation (E) step. For the maximization (M) step, the first derivatives are
simply set to zero:
L x ( )
=
j
(k )
A
A
= 0
ij
j
ij
(k )
j
Aij j + bi
yi
(13.66)
This is easily solved for j, yielding the new reconstruction j(k+1) :

j(k +1) =
j(k )
yi
A
A
Aij
ij
ij
(k )
j
+ bi
(13.67)
This is the well known MLEM algorithm for emission tomography.

It can be shown that this recipe has the wonderful feature that each new
EM iteration increases the value of the likelihood LML. It should be noted that the
complete data xij do not appear in Eq.(13.67); they are needed in the derivation
but they do not need to be computed explicitly. This is very fortunate as there is a
huge number of them.
An initial image (1) is required to start the iterations. As experience (and
theoretical analysis) has shown that higher spatial frequencies have slower
convergence, and because smooth images are preferred, the initial image is
usually chosen to be uniform, by setting j(1)=C and j=1J, where C is a
strictly positive constant.
The MLEM algorithm is multiplicative, implying that it cannot change the
value of a reconstruction voxel, when the current value is zero. For this reason,
the voxels in the initial image should only be set to zero if it is known a priori that
they are indeed zero. The derivation of the MLEM algorithm uses the assumption
that all yi, all xij and all j are non-negative. Assuming that yi 0 and i=1I,
and considering that the probabilities Aij are also non-negative, it is clear that
when the initial image (1) is non-negative, all subsequent images (k) will be
non-negative as well. However, when, for some reason, a reconstruction value
becomes negative (e.g. because one or a few sinogram values yi are negative),
then convergence is no longer guaranteed. In practice, divergence is almost
guaranteed in that case. Consequently, if the sinogram is pre-processed with a
procedure that may produce negatives (e.g. randoms subtraction in PET), MLEM
reconstruction will only work if all negative values are set to a non-negative
value.
480
13.3.3.2. Reconstruction from list-mode data

The measured data yi considered in the derivations above (so-called binned
data) represent the number of counts acquired within an individual crystal pair i
(LOR i), that is, yi represents the sum of those acquired events (indexed by m)
that were assigned (histogrammed) to the i-th LOR: y i = mi 1. However, in
modern PET systems, the number of possible LORs within the FOV typically
exceeds (often by many times) the number of events acquired in a clinical PET
study. Consequently, the binned data are very sparse and it is more efficient
to store and process each acquired event (with all of its relevant information)
separately, in the so-called list-mode format.
Modification of the maximum-likelihood algorithms is straightforward
(whether MLEM or accelerated algorithms based on ordered subsets discussed
later), as shown in works by Parra and Barrett [13.22], and by Reader et al. [13.23].
It should be noted that the same is not true about other algorithms, for example,
algorithms with additive updates. The MLEM algorithm for the list-mode data
can be obtained by replacing yi in the MLEM equation (Eq.(13.67)) by the above
mentioned sum over events, skipping the LORs with zero counts (which do not
contribute to the MLEM sum), and combining the sum over LORs i with the sum
over events m:
j(k +1) =
j(k )
iLORs
Aij
mevent-list
Ai
mj
A
j
(k )
i m j j + bi m
(13.68)
where im represents the LOR index in which the m-th event has been recorded.
The main difference is that the MLEM sum is now evaluated (including
calculations of the relevant forward and back projections) only over the list of
the available events (in any order). However, it is important to mention here that
the normalizing term in front of the sum (sensitivity matrix
A ) still has to
i ij
be calculated over all possible LORs, and not only those with non-zero counts.
This represents a challenge for the attenuated data (attenuation considered as
part of the system matrix A), since the sensitivity matrix has to be calculated
specifically for each object and, therefore, it cannot be pre-computed. For modern
systems with a large number of LORs, calculation of it often takes more time
than the list-mode reconstruction itself. For this reason, alternative approaches
(involving certain approximations) have been considered for the calculation of
the sensitivity matrix, such as subsampling approaches [13.24] or Fourier based
approaches [13.25].
481
CHAPTER 13
13.3.3.3. Reconstruction of time of flight PET data

In the TOF case, the probability of a pair of photons arriving from a particular
point along the LOR (as reported based on the difference of their detection times) is
given by a Gaussian kernel having a width determined by the timing uncertainty of
the detection system. In contrast, in the non-TOF case, the probability of detecting
the event is approximately uniform along the LOR. Modification of iterative
reconstruction algorithms (whether for binned or list-mode data) to account for the
TOF is straightforward. Integrations along the LORs (the main component of the
system matrix A) just need to be replaced with the TOF kernel weighted integrations
along the LORs. The forward projection (or back projection) in a certain direction
can now be viewed, and performed, as a convolution of the image with a proper
TOF kernel in the LOR direction (see Fig. 13.13). The rest of the algorithm, i.e.
formulas derived in the previous subsections, stays exactly the same (only the form
of the system matrix A is changed). Additional information provided by the TOF
measurements, leading to more localized data, results in faster, and more uniform,
convergence, as well as in improved signal to noise ratios in reconstructed images,
as widely reported in the literature.
Projection
(LOR-binned events)
Projection
(LOR-binned events)
Histo-Image
Histo-Projection
(image-binned events)
(LOR & TOF-binned events)

histo-image voxels
image voxels
histo-projection bins =
TOF-extended projection bins
FIG. 13.13. Comparison of the data formats for binned time of flight (TOF) data
(left: histo-projection for a 45 view) and for the DIRECT (direct image reconstruction for
TOF) approach (right: histo-image for a 45 view). Histo-projections can be viewed as an
extension of individual non-TOF projections into TOF directions (time bins), and their
sampling intervals relate to the projection geometry and timing resolution. Histo-images are
defined by the geometry and desired sampling of the reconstructed image. Acquired events
and correction factors are directly placed into the image resolution elements of individual
histo-images (one histo-image per view) having a one to one correspondence with the
reconstructed image voxels.
The TOF mode of operation has some practical consequences (and novel
possibilities) for the ways the acquired data are stored and processed. The
482
list-mode format is very similar to the non-TOF case. The event structure is just
slightly expanded by a few bits (58 bits/event) to include the TOF information,
and the events are processed event by event as in the non-TOF case.
On the other hand, the binned data undergo considerable expansion when
accommodating the TOF information. The projection (X ray transform) structures
are expanded by one dimension, that is, each projection bin is expanded in the
LOR direction into the set of time bins forming the so-called histo-projections
(seeFig.13.13 (left)). In practice, the effect of this expansion on the data size is
not as bad as it appears, because the localized nature of TOF data allows decreased
angular sampling (typically about 510 times) in both azimuthal and co-polar
directions (views), while still satisfying angular sampling requirements. The
resulting data size, thus, remains fairly comparable to the non-TOF case. During
the reconstruction process, the histo-projection data are processed time-bin
by time-bin (instead of projection line by line in the non-TOF case). It should
be noted that hybrid approaches also exist between the two aforementioned
approaches, in which the data are binned in the LOR space, but events are stored
in list-mode for each LOR bin.
TOF also allows a conceptually different approach of data partitioning,
leading to more efficient reconstruction implementations, by using the DIRECT
(direct image reconstruction for TOF) approach utilizing so-called histo-images
(seeFig.13.13 (right)) [13.25]. In the DIRECT approach, the data are directly
histogrammed (deposited), for each view, into image resolution elements
(voxels) of desired size. Similarly, all correction arrays and data are estimated
or calculated in the same histo-image format. The fact that all data and image
structures are now in image arrays (of the same geometry and size) makes
possible very efficient computer implementations of the data processing and
reconstruction operations.
13.3.3.4. Reconstruction of dynamic data
Data acquired from an object dynamically changing with time in activity
distribution, or in morphology (shape), or in both is referred to as dynamic data.
An example of the first case would be a study looking at temporal changes in
activity uptake in individual organs or tissues, so-called timeactivity curves. An
example of the second case would be a gated cardiac study providing information
about changes of the heart morphology during the heart beat cycle (such as
changes of the heart wall thickness or movements of the heart structures).
The dynamic data can be viewed as an expansion of static (3D) data
by the temporal information into 4D (or 5D) data. The dynamic data are
usually subdivided (spread) into a set of temporal (time) frames. In the
first application, each time frame represents data acquired within a certain
483
CHAPTER 13
sequential time subinterval of the total acquisition time. The subintervals can
be uniform, or non-uniform with their durations adjusted, for example, to the
speed of the change of the activity curves. In the second application, each time
frame represents the total counts acquired within a certain stage (gate) of the
periodic organ movement (e.g. gated based on the electrocardiogram signal).
In the following, issues of the reconstruction of dynamic data in general are
addressed. Problems related to the motion and its corrections are discussed in
Section 13.3.6.4.
Once the data are subdivided (during acquisition) or sorted (acquired
list-mode data) into the set of time frames, seemingly the most natural way of
reconstructing them is to do it for each time frame separately. It should be noted
that this is the only available option for the analytical reconstruction approaches,
while the iterative reconstruction techniques can also reconstruct the dynamic
data directly in 4D (or 5D). A problem with frame by frame reconstruction is
that data in the individual time frames are quite noisy, since each time frame
only has a fraction of the total acquired counts, leading to noisy reconstructions.
Consequently, the resulting reconstructions often have to be filtered in the spatial
and/or temporal directions to obtain images of any practical value. Temporal
filtering takes into account time correlations between the signal components in
the neighbouring time frames, while the noise is considered to be independent.
Filtering, however, leads to resolution versus noise trade-offs.
On the other hand, reconstructing the whole 4D (or 5D) dataset together,
while using this correlation information in the (4D) reconstruction process via
proper temporal (resolution) kernels or basis functions, can considerably improve
those trade-offs as reported in the literature (similarly to the case of spatial
resolution modelling). The temporal kernels (basis functions) can be uniform in
shape and distribution, or can have a non-uniform shape (e.g. taking into account
the expected or actual shape of the timeactivity curves) and can be distributed
on a non-uniform grid (e.g. reflecting count levels at individual frames or image
locations). The kernel shapes and distributions can be defined, or determined,
beforehand and be fixed during the reconstruction. During the reconstruction
process, just the amplitudes of the basis functions are reconstructed. The
algorithms derived in the previous subsections basically stay the same, where the
temporal kernels can be considered as part of the system matrix A (comparable
to including the TOF kernel in TOF PET). Another approach, more accurate but
mathematically and computationally much more involved, is to iteratively build
up the shape (and distribution) of the temporal kernels during the reconstruction
in conjunction with the reconstruction of the emission activity (that is, the
amplitude of the basis functions).
While iterative methods lead to a clear quality improvement when
reconstructing dynamic data, thanks to the more accurate models of the signal and
484
data noise components, for the quantitative dynamic studies their shortcoming
is their non-linear behaviour, especially if they are not fully converged. For
example, the local bias levels can vary across the time frames as the counts, local
activity levels and object morphology change, which can lead to less accurate
timeactivity curves. On the other hand, analytical techniques which are linear
and consequently do not depend on the count levels and local activity, might
provide a more consistent (accurate) behaviour across the time frames in the
mean (less bias of the mean), but much less consistent (less precise) behaviour in
the variance due to the largely increased noise. It is still an open issue which of
the two approaches provides more clinically useful results, and the discussions
and research on this topic are still open and ongoing.
13.3.4. Acceleration
13.3.4.1. Ordered-subsets expectation-maximization
The MLEM algorithm requires a projection and a back projection in
every iteration, which are operations involving a large number of computations.
Typically, MLEM needs several tens to hundreds of iterations for good
convergence. Consequently, MLEM reconstruction is slow and many researchers
have studied methods to accelerate convergence.
The method most widely used is ordered-subsets expectation-maximization
(OSEM) [13.26]. The MLEM algorithm (Eq.(13.67)) is rewritten here for
convenience:
y i(k ) =
(13.69)
(k )
ij j + bi
j(k +1) =
j(k )
A
A
ij
ij
yi
y i(k )
(13.70)
where k is the iteration number and (1) is typically set to a uniform, strictly
positive image.
In OSEM, the set of all projections {1 ... I} is divided into a series of subsets
St, t=1T. Usually, these subsets are exhaustive and non-overlapping, i.e. every
projection element i belongs to exactly one subset St. In SPECT and PET, the data
y are usually organized as a set of (parallel- or fan-beam) projections, indexed
by projection angle . Therefore, the easiest way to produce subsets of y is by
assigning all of the data for each projection angle to exactly one of the subsets.
485
CHAPTER 13
However, if the data y are stored in list-mode (seeSection 13.3.2), the easiest way
is to simply cut the list into blocks, assigning each block to a different subset.
The OSEM algorithm can then be written as:
initialize jold , j=1,J
for k=1,K
for t=1,T
y i =
old
ij j
+ bi ,
i St
for j=1,J
old
jnew = j
A
A
iS t
ij
ij iS t
yi
(13.71)
y i
If all of the projections are combined into a single subset, the OSEM algorithm is
identical to the MLEM algorithm. Otherwise, a single OSEM iteration k consists
of T sub-iterations, where each sub-iteration is similar to an MLEM iteration,
except that the projection and back projection are only done for the projections of
the subset St. If every sinogram pixel i is in exactly one subset, the computational
burden of a single OSEM iteration is similar to that of an MLEM iteration.
However, MLEM would update the image only once, while OSEM updates it T
times. Experience shows that this improves convergence by a factor of about T,
which is very significant.
Convergence is only guaranteed for consistent data and provided that there
is subset balance, which requires:
A =A
ij
iS t
ij
(13.72)
iS u
where St and Su are different subsets.

In practice, these conditions are never satisfied, and OSEM can be shown to
converge to a limit cycle rather than to a unique solution, with the result that the
OSEM reconstruction is noisier than the corresponding MLEM reconstruction.
However, in many applications, the difference between the two is not clinically
relevant.
The procedure is illustrated with a simple simulation in Fig.13.14. As there
was no noise and no attenuation, convergence of OSEM is guaranteed in this
example. In more realistic cases, it may be recommended to have four or more
486
projections in a single subset, to prevent excessive noise amplification at higher

iteration numbers.
MLEM iterations
FIG.13.14. A simulation comparing a single ordered-subsets expectation-maximization
(OSEM) iteration with 40 subsets, to 40 maximum-likelihood expectation-maximization
(MLEM) iterations. The computation time of the MLEM reconstruction is about 40 times
longer than that of OSEM. In this example, there were only two (parallel-beam) projection
angles per subset, which is clearly visible in the first OSEM iteration.
13.3.4.2. Refinements of the ordered-subsets expectation-maximization algorithm

As mentioned above, OSEM converges to a limit cycle: after many
iterations, it starts cycling through a series of solutions rather than converging to
the maximum-likelihood solution. When compared to the initial image (usually a
uniform image), these series of solutions are relatively close to the maximumlikelihood solution. Consequently, the convergence of OSEM is initially much
faster but otherwise similar to that of MLEM; the better performance of MLEM
only becomes noticeable at high iteration numbers. Thus, a simple solution to
avoid the limit cycle is to gradually decrease the number of subsets: this approach
preserves the initial fast convergence of OSEM, avoiding the limit cycle by
returning to MLEM at high iteration numbers. A drawback of this approach is
that convergence becomes slower each time the number of subsets is reduced.
In addition, there is no theory available that prescribes how many sub-iterations
should be used for each OSEM iteration.
Many algorithms have been proposed that use some form of relaxation to
obtain convergence under less restrictive conditions than those of OSEM. As an
example, relaxation can be introduced by rewriting the OSEM Eq.(13.71) in an
487
CHAPTER 13
additive way. Then, a relaxation factor is inserted to scale the update term to
obtain RAMLA (row-action maximum-likelihood algorithm [13.27]):
jnew = jold + jold
A y
ij
iS t
i
i
with
<
max t (
A )
ij
(13.73)
iS t
The relaxation factor decreases with increasing iteration number to ensure
convergence. It should be noted that setting = 1 / Aij for all (sub-)iterations

iS t
yields OSEM. Several alternative convergent block iterative algorithms have

been proposed. They are typically much faster than MLEM but slightly slower
than the (non-convergent) OSEM algorithm.
13.3.5. Regularization
MLEM maximizes the likelihood, by making the computed projections
(from the current reconstruction) as similar as possible to the measured projections,
where the similarity is measured based on the Poisson distribution. An upper limit
of the likelihood would be obtained when the measured and calculated projections
are identical. However, this is never possible, because Poisson noise introduces
inconsistencies. Nevertheless, a large part of the noise is consistent, which means
that it can be obtained as the projection of a (noisy) activity distribution. This part
of the noise propagates into the reconstructed image, and is responsible for the
so-called deterioriation of the MLEM image at high iterations.
13.3.5.1. Stopping iterations early
An accidental feature of the MLEM algorithm is its frequency dependent
convergence: low spatial frequencies converge faster than higher frequencies.
This is due to the low-pass effect of the back projection operation. This effect is
easily verified for the reconstruction of the activity in a point source, if the MLEM
reconstruction is started from a uniform image. The first iteration then yields the
back projection of the point source measurement. As discussed in Section 13.2.1,
this yields an image with intensity ( x, y) 1 / x 2 + y 2 , if the point source was
located at (0,0). Each iteration multiplies with a similar back projection, implying
that after t iterations, the image intensity at (x,y) is proportional to 1/(x2 + y2)t/2,
so that the peak at (0,0) becomes a bit sharper with every iteration. For more
complicated objects, the evolution is more subtle.
488
True image
Sinogram
Smoothed
true image
With
noise
Smoothed
FIG.13.15. Simulation study illustrating position dependent convergence in PET with

attenuation. After 8 iterations (iter), convergence in highly attenuated regions is poor.
After 100 iterations, good convergence is obtained, but with strong noise propagation.
Post-smoothing yields a fair compromise between noise and nearly position independent
resolution. FBP: filtered back projection.
It follows that reducing the number of iterations has an effect which is

similar to reducing the cut-off frequency of a low-pass filter. However, the effect
on the resolution is position dependent, as illustrated in Fig.13.15. Attenuated
PET projections of a highly radioactive uniform ring inside a less active disc
were simulated with and without Poisson noise. After eight MLEM iterations,
the reconstructed ring has non-uniform activity. In the centre of the phantom,
convergence is slower, resulting in poorer resolution and poorer recovery of the
activity in the ring. After 100 iterations, convergence is much better everywhere
in the phantom, but for noisy data, there is very disturbing noise propagation.
If the image was acquired for detection (e.g. to see if there is a radioactive
ring inside the disc or not), then the image produced after eight iterations is
excellent. However, if the aim is quantification (e.g. analysing the activity
distribution along the ring), then quantification errors can be expected at low
iteration numbers.
13.3.5.2. Post-smoothed maximum-likelihood
The noise in the higher MLEM iterations is high frequency noise, and
there are strong negative correlations between neighbouring pixels. As a result, a
489
CHAPTER 13
modest amount of smoothing strongly suppresses the noise at the cost of a mild
loss of resolution. This is illustrated in the third row of Fig.13.15.
If the MLEM implementation takes into account the (possibly position
dependent) spatial resolution effects, then the resolution should improve with
every MLEM iteration. After many iterations, the spatial resolution should be
rather good, similar or even better than the sinogram resolution, but the noise
will have propagated dramatically. It is assumed that the obtained spatial
resolution corresponds to a position dependent point spread function which can
be approximated as a Gaussian with a full width at half maximum (FWHM)
of FML(x, y). Assume further that this image is post-smoothed with a (position
independent) Gaussian convolution kernel with an FWHM of Fp. The local
point spread function in the smoothed image will then have an FWHM of
(F ML ( x, y)) 2 Fp2 . If enough iterations are applied and if the post-smoothing
kernel is sufficiently wide, the following relation holds Fp Fml(x, y) and,
therefore, (F ML ( x, y)) 2 + Fp2 Fp . Under these conditions, the post-smoothed
MLEM image has a nearly position independent and predictable spatial
resolution. Thus, if PET or SPECT images are acquired for quantification, it is
recommended to use many iterations and post-smoothing, rather than a reduced
number of iterations, for noise suppression.
13.3.5.3. Smoothing basis functions
An alternative approach to counter noise propagation is to use an image
representation that does not accomodate noisy images. Instead of representing the
image with a grid of non-overlapping pixels, a grid of smooth, overlapping basis
functions can be used. The two mostly used approaches are the use of spherical
basis functions or blobs [13.28] and the use of Gaussian basis functions or
sieves [13.29].
In the first approach, the projector and back projector operators are
typically adapted to work directly with line integrals of the basis functions. In
the sieves approach, the projection of a Gaussian blob is usually modelled as the
combination of a Gaussian convolution and projection along lines. The former
approach produces a better approximation of the mathematics, while the latter
approach yields a faster implementation.
The blobs or sieves are probably most effective when their width is very
similar to the spatial resolution of the tomographic system. In this setting, the basis
function allows accurate representation of the data measured by the tomographic
system, and prevents reconstruction of much of the (high frequency) noise. It has
been shown that using the blob during reconstruction is more effective than using
the same blob only as a post-smoothing filter. The reason is that the post-filter
490
always reduces the spatial resolution, while a sufficiently small blob does not
smooth data if it is used during reconstruction.
If the blob or sieve is wider than the spatial resolution of the tomographic
system, then its use during reconstruction produces Gibbs over- and undershoots,
also known as ringing. This effect always arises when steep edges have to be
represented with a limited frequency range, and is related to the ringing effects
observed with very sharp low-pass filters. For some imaging tasks, these ringing
artefacts are a disadvantage.
13.3.5.4. Maximum a posteriori or penalized likelihood
Smoothing the MLEM image is not a very elegant approach: first, the
likelihood is maximized, and then it is decreased again by smoothing the image.
It seems more elegant to modify the objective function, such that the image that
maximizes it does not need further processing. This can be done with a Bayesian
approach, which is equivalent to combining the likelihood with a penalty
function.
It is assumed that a good reconstruction image will be obtained if that
image maximizes the (logarithm of the) probability p(|y) given by Eq.(13.39)
and repeated here for convenience:
= arg max(ln p( y | ) + ln p( )) (13.74)
The second term represents the a priori knowledge about the tracer distribution,
and it can be used to express our belief that the true tracer distribution is fairly
smooth. This is usually done with a Markov prior. In a Markov prior, the a priori
probability for a particular voxel, given the value of all other voxels, is only a
function of the direct neighbours of that voxel:
p( j | k , k j ) = p( j | k , k N j ) (13.75)
where Nj denotes the set of neighbour voxels of j.

Such priors are usually written in the following form:
P( ) = ln p( ) =
ln p(
j
where
| k , k N j ) =
E(
j k ) (13.76)
k N j
the energy function E is designed to obtain the desired noise suppressing

behaviour and the parameter is the weight assigned to the prior.
491
CHAPTER 13
A higher weight results in smoother images, at the cost of a decreased

likelihood, i.e. poorer agreement with the acquired data. In most priors, the
expression is further simplified by making E a function of a single variable, the
absolute value of the difference |j k|.
Some popular energy functions E(|j k|) are shown in Fig.13.16. A simple
and effective one is the quadratic prior E(x)=x2; an MAP reconstruction with
this prior is shown in Fig.13.17. Better preservation of strong edges is obtained
with the Huber prior: it is quadratic for |j k| and linear for |j k| > ,
with a continuous first derivative at . Consequently, it applies less smoothing
than the quadratic prior for differences larger than , as illustrated in Fig.13.17.
Even stronger edge tolerance is obtained with the Geman prior, which converges
asymptotically to a constant for large differences, implying that it does not
smooth at all over very large pixel differences.
Quadratic
Huber
Geman
FIG.13.16. The energy function of the quadratic prior, the Huber prior and the Geman prior.
Original
MLEM
Quadratic
Huber
Geman
FIG. 13.17. Maximum-likelihood expectation-maximization (MLEM) and maximum a

posteriori reconstructions of the SheppLogan phantom. Three different smoothing priors were
used: quadratic, Huber and Geman. The latter smooth small differences quadratically, but are
more tolerant for large edges.
It can be shown that the prior (Eq.(13.76)) is a concave function of if

E|j k| is a convex function. Consequently, the quadratic and Huber energy
492
functions yield a concave prior: it has a single maximum. In contrast, the Geman
prior is not concave (seeFig.13.16) and has local maximums. Such concave
priors require careful initialization, because the final reconstruction depends on
the initial image and on the behaviour of the optimization algorithm.
Figure13.18 shows that MAP reconstructions produce position dependent
spatial resolution, similar to MLEM with a reduced number of iterations. The
reason is that the prior is applied with a uniform weight, whereas the likelihood
provides more information about some voxels than about others. As a result,
the prior produces more smoothing in regions where the likelihood is weaker,
e.g. regions that have contributed only a few photons to the measurement due to
high attenuation.
MLEM
Smoothed MLEM
MAP quadratic prior
FIG.13.18. Maximum-likelihood expectation-maximization (MLEM), smoothed MLEM and

maximum a posteriori (MAP) (quadratic prior) reconstructions of simulated PET data of a
brain and a ring phantom. The ring phantom reveals position dependent smoothing for MAP.
The prior can be made position dependent as well, to ensure that the balance
between the likelihood and the prior is about the same in the entire image. In
that case, MAP with a quadratic prior produces images which are very similar
to MLEM images with post-smoothing: if the prior and smoothing are tuned to
produce the same spatial resolution, then both algorithms also produce nearly
identical noise characteristics.
493
CHAPTER 13
Many papers have been devoted to the development of algorithms for MAP
reconstruction. A popular algorithm is the so-called one step late algorithm.
Inserting the derivative of the prior P in Eq.(13.66) yields:
(L x ( ) + P( ))
=
j
y
i
y
i
(k )
i
Aij j(k )
P( )
Aij +
= 0 (13.77)
j
j
where y i( k ) is the projection of the current reconstruction for detector i.

A problem with this equation is that P( ) / j is itself a function of the
unknown image . To avoid this problem, the derivative of the prior is simply
evaluated in the current reconstruction (k). The equation can then be solved to
produce the MAP update expression:
j(k +1)
j(k )
P( )
Aij
j
ij
yi
y i(k )
(13.78)
(k )
Owing to the approximation, convergence is not guaranteed. The algorithm

usually works fine, except with very high values for the prior.
The MLEM algorithm can be considered as a gradient ascent algorithm
(seealso Eq.(13.50)):
j(k +1) =
j(k )
ij
Aij
yi
(13.79)
y i(k )
= (j k ) +
(j k ) L ML ( )
j
Aij
(13.80)
(k )
Extensions to an MAP gradient ascent algorithm typically have the form:

j(k +1) = j(k ) + S( (k ) )
(LML ( ) + P( ))
j
(13.81)
(k )
where the key is to determine a good preconditioner S.
494
Several methods with (almost) guaranteed convergence have been based

on the previously described optimization transfer method, by designing useful
surrogate functions for both the likelihood and the prior.
13.3.6. Corrections
In typical emission data, the true events (having a Poisson character) are
distorted and contaminated by a number of physical effects. To make the best use
of the acquired data and of our knowledge of the acquisition system, these effects
should be included in the reconstruction model. The distortion effects include
resolution effects (such as detector resolution, collimator effects, and in PET
also non-collinearity and positron range) and motion effects. The contamination
effects can be divided, by their character and the way they are treated, into
multiplicative and additive terms. The multiplicative factors include: attenuation
of the annihilation photons by the object, the probability of the detector elements
detecting an event once they are hit by the photon (detector normalization
factors), coefficients accounting for the decay time and the geometrical restriction
of directions/LORs for which true events are detected (axial acceptance angle,
detector gaps). The additive terms include scattered and random (in the PET
case) coincidences. Details on calculation of the correction factors and terms
are discussed in other chapters. This chapter is limited to the discussion of their
utilization within the reconstruction process.
The most straightforward approach is to pre-correct the data before
reconstruction for the contamination effects (multiplying by multiplicative
correction coefficients and subtracting the scatter and random estimates),
so as to approximate the X ray transform (or attenuated X ray transform in
the SPECT case) of the reconstructed object. For analytical reconstruction
approaches (derived for the ideal X ray transform data), the data always have to
be pre-corrected.
For the statistical reconstruction methods, derived based on the statistical
properties of the data, an attempt is made to preserve the Poisson character
of the data as much as possible by including the correction effects inside the
reconstruction model. Theoretically, the most appropriate way is to include the
multiplicative and scatter effects directly into the system matrix. The system
matrix would have to include not only an accurate model of the direct data (true
events) but also of the physical processes of the generation of the contamination
scatter data. In a sense, the contamination would then become valid data, bringing
extra information to our model and, thus, adding valid (properly modelled)
counts to the image. However, inclusion of the scatter model into the system
matrix tremendously increases the number of non-zero elements of the system
matrix, i.e. the matrix is not sparse anymore, and consequently the system is more
495
CHAPTER 13
ill-posed (the contamination data are typically quite noisy) and computationally
exceedingly expensive, and, thus, not feasible for routine clinical use.
The more practical, and commonly used, approach is to include correction
effects as multiplicative factors and additive terms within the forward projection
model of the iterative reconstruction approaches:
y=A + b
(13.82)
where the effects directly influencing the direct (true) data are included inside the
system matrix A and will be discussed in the following, while the additive terms
b (including scatter and randoms) will be discussed separately in Section 13.3.6.2
on additive terms.
13.3.6.1. Factors affecting direct events multiplicative effects
In the PET case, the sequence of the physical effects (described in previous
chapters) that occur as the true coincident events are generated and detected can
be described by the following factorization of the system matrix A as discussed
in detail in Ref. [13.30]:
A = Adet.sens Adet.blur Aatt Ageom Atof Apositron (13.83)
where
Apositron models the positron range;
Atof models the timing accuracy for the TOF PET systems (TOF resolution
effects, as discussed in Section 13.3.3.3);
Ageom
is the geometric projection matrix, the core of the system matrix,
which is a geometrical mapping between the source (voxel j) and data
(projection bin i, defined by the LOR, or its time bin in the TOF case);
the geometrical mapping is based on the probability (in the absence of
attenuation) that photon pairs emitted from an individual image location
(voxel) reach the front faces of a given crystal pair (LOR);
Aatt is a diagonal matrix containing attenuation factors on individual LORs;
Adet.blur
models the accuracy of reporting the true LOR positions (detector
resolution effects; discussed in Section 13.3.6.2);
and Adet.sens is a diagonal matrix modelling the probability that an event will
be reported once the photon pair reaches the detector surface a unique
multiplicative factor for each detector crystal pair (LOR) modelled by
496
normalization coefficients, but can also include the detector axial extent and
detector gaps.
In practice, the attenuation operation Aatt is usually moved to the left (to
be performed after the blurring operation). This is strictly correct only if the
attenuation factors change slowly, i.e. they do not change within the range
of detector resolution kernels. However, even if this is not the case, a good
approximation can be obtained by using blurred (with the detector resolution
kernels) attenuation coefficients. In this case, the multiplicative factors Adet.sens
and Aatt can be removed from the system matrix A and applied only after the
forward projection operation as a simple multiplication operation (for each
projection bin). The rest of the system matrix (except Apositron, which is object
dependent) can now be pre-computed, whether in a combined or a factorized
form, since it is now independent of the reconstructed object. On the other hand,
the attenuation factors Aatt (and Apositron, if considered) have to be calculated for
each given object.
In the SPECT case, the physical effects affecting the true events can be
categorized and factorized into the following sequence:
A = Adet.sens Adet.blur Ageom,att (13.84)
where
Adet.sens includes multiplicative factors (such as detector efficiency and decay
time);
Adet.blur represents the resolution effects within the gamma camera (the intrinsic
resolution of the system);
and Ageom,att is the geometric projection matrix, also including the collimator
effects (such as the depth dependent resolution) and the depth and view dependent
attenuation factors.
For gamma cameras, the energy and linearity corrections are usually
performed in real time, and the remaining (detector efficiency) normalization
factors are usually very close to one and can be, for all practical purposes, ignored
or pre-corrected. Similarly, the theory says that the decay correction should be
performed during the reconstruction, because it is different for each projection
angle. However, for most tracers, the decay during the scan is very modest,
and in practice it is usually either ignored or done as a pre-correction. The
attenuation component is object dependent and needs to be recalculated for each
reconstructed object. Furthermore, its calculation is much more computationally
497
CHAPTER 13
expensive than in the PET case, since it involves separate calculations of the
attenuation factors for each voxel and for each view. This is one of the reasons
why the attenuation factors have often been ignored in SPECT. More details on
the inclusion of the resolution effects into the system matrix are discussed in
Section 13.3.6.3.
13.3.6.2. Additive contributions
The main additive contaminations are scatter (SPECT and PET) and
random events (PET). The simplest possibility of dealing with them is to
subtract their estimates ( s and r ) from the acquired data. While this is a valid
(and necessary) pre-correction step for the analytical reconstructions, it is not
recommended for statistical approaches since it changes the statistical properties
of the data, causing them to lose their Poisson character. As the maximumlikelihood algorithm is designed for Poisson distributed data, its performance is
suboptimal if the data noise is different from Poisson. Furthermore, subtraction of
the estimated additive terms from the noisy acquired data can introduce negative
values into the pre-corrected data, especially for low count studies. The negative
values have to be truncated before the maximum-likelihood reconstruction, since
it is not able to correctly handle the negative data. This truncation, however, leads
to a bias in the reconstruction.
On the other end of the spectrum of possibilities, would be considering
the scatter and randoms directly in the (full) system model, that is, including a
complete physical model of the scatter and random components into a Monte
Carlo calculation of the forward projection. However, this approach is exceedingly
computationally expensive and is not feasible for practical use. A practical and
the most common approach for dealing with the additive contaminations is to add
their estimate (b = s + r ) to the forward projection in the matrix model of the
iterative reconstruction, i.e. the forward model is given by A + b , as considered
in the derivation of the MLEM reconstruction (Eq.(13.67)).
Special treatment has to be considered for clinical scanners in which the
random events (r, estimated by delayed coincidences) are on-line subtracted
from the acquired data (y, events in the coincidence window prompts).
The most important characteristic of the Poisson data is that their mean equals
their variance: mean(yi)=var(yi). However, after the subtraction of the delays
from the prompts (both being Poisson variables), the resulting data () are
not Poisson anymore, since mean(i) = mean(yi ri) = mean(yi) mean(ri),
while var(i) = var(yi ri) = var(yi) + var(ri). To regain the main characteristic
of the Poisson data (at least of the first two moments), the shifted Poisson
approach can be used, utilizing the fact that adding a (noiseless) constant
value to the Poisson variable changes the mean but preserves the variance
498
of the result. To modify the mean of the subtracted data to be equal to

their variance (i.e. var(yi) + var(ri)), we need to add to the subtracted data an
estimate (of the mean) of the randoms (r ) multiplied by two. This gives
mean(g i + 2 ri ) = mean( y i ri + 2 ri ) = mean( y i ) + mean(ri ) , which is equal to
var(g i + 2 ri ) = var( y i ) + var ( y i ) . The MLEM algorithm using the shifted Poisson
model can then be written as:
j(k +1) =
j(k )
i + 2 ri
A
A A
ij
ij
ij
(k )
j + si
+ 2 ri
(13.85)
It is worthwhile mentioning here that even in the shifted Poisson case, the
negative values in the subtracted data and consequent truncation leading to the
bias and artefacts cannot be completely avoided. However, the chance of the
negative values decreases since the truncation of the negative values is being
performed on the value-shifted data ( i + 2ri ) . Examples of reconstructions
from data with a subtracted additive term, using the regular MLEM algorithm
and using MLEM with the shifted Poisson model, are shown in Fig.13.19. As the
counts were relatively high in this simulation, the subtraction did not produce
y
Original
Original +
contaminator
yr
MLEM of (y r)
MLEM of
(y r + 2r, 2r)
FIG.13.19. Illustration of (exaggerated case of) reconstructions from contaminated data y

from which the additive contamination term r was subtracted (both data and contamination
term are Poisson). The top row shows the sinograms. The increased noise level in the
contaminated area in the sinogram (y r) should be noted. The bottom row shows the true
image without and with the contaminator, the maximum-likelihood expectation-maximization
(MLEM) reconstruction from the subtracted data (y r) and the shifted Poisson MLEM
reconstruction, in which the estimated (noiseless) additive term 2r is added to the subtracted
data and forward projection as given by Eq.(13.85).
499
CHAPTER 13
negatives. MLEM of (y r) creates streaks because the reliability of the subtracted

data is overestimated.
It should be noted that in the reconstruction model (as well as in the
pre-correction approaches) the estimates of the scatter and randoms have to
be treated in the same way as the estimates of the true events in the forward
projection, including consideration of the normalized or un-normalized events,
attenuation corrected or uncorrected data, gaps in the data, etc. Various challenges
exist for the scatter and randoms estimations in general, such as modelling of the
out of FOV scatter. This is addressed in Chapter11.
13.3.6.3. Finite spatial resolution
There are a number of physical and geometrical effects and limitations
(such as positron range, non-collinearity, depth of interaction, size of detector
crystal elements, inter-crystal scatter, collimator geometry, etc.) affecting PET
and SPECT resolution as described in more detail in Chapter11. To get the most
out of the acquired data and to correct for the resolution degradation, these effects
have to be properly modelled in the system matrix of statistical reconstruction,
as considered in the components (Adet.blur, Ageom, Apositron) of the factorized
system matrix outlined in Section 13.3.6.1. This step does not influence the
mathematical definition of the reconstruction algorithm (such as MLEM, as
given by Eq.(13.67)); only the form of its system matrix is changed.
However, this step has very practical consequences for the complexity of
the algorithm implementation, for computational demands and most importantly
for the quality of the reconstructed images. By including the resolution effects
into the reconstruction model, a larger fraction of the data is being used for the
reconstruction within each point of the space, with the true signal component
becoming more consistent, while the noise components becoming less consistent
with the model. Thus, the resolution modelling helps twice, by improving the
image resolution while at the same time reducing the image noise, as illustrated
in Fig.13.20 for simulated SPECT data. This is quite different from the
filtering case, where the noise suppression is always accompanied by resolution
deterioration. On the other hand, the resolution modelling has a price in terms
of a considerable increase in the computational load (both in space/memory and
time) since the system matrix is much less sparse, that is, it contains a larger
proportion of non-zero elements. This not only leads to more computational load
per iteration, but also to a slower convergence of the iterative reconstruction and,
consequently, to the need for more iterations.
Resolution effects can be subdivided into the effects dependent on the
particular object, such as the positron range, and the effects influenced by the
scanner geometry, design and materials (which can be determined beforehand
500
for the given scanner). The positron range depends on the particular attenuation
structures in which the the positrons annihilate, and also varies from isotope
to isotope. Furthermore, the shape of the probability function (kernel) of the
positron annihilation abruptly changes at the boundaries of two tissues, such as
at the boundary of the lungs and surrounding soft tissues, and, thus, it strongly
depends on the particular objects morphology and is quite challenging to model
accurately. In general, the positron range has a small effect (compared to the other
effects) for clinical scanners, particularly for studies using 18F-labelled tracers,
and can often be ignored. However, for small animal imaging and for other tracers
(such as 82Rb), the positron range becomes an important effect to be considered.
Simulated SPECT data
No
Poisson noise
MLEM resolution model
MLEM
MLEM
FIG.13.20. Examples of the effects of resolution modelling within statistical iterative

reconstruction. Data were simulated for a SPECT system with depth dependent resolution. It is
clearly seen that using the proper resolution model within statistical reconstruction (lower two
images on the right) not only improves the resolution of the images, but also helps to efficiently
suppress the noise component.
There is a whole spectrum of approaches to determine and implement the

scanner dependent resolution models. Only the main ones are addresssed. The
simplest, but least accurate, approach is to approximate the system resolution
model by a spatially invariant resolution kernel, usually a spherically symmetric
Gaussian, with the shape (FWHM) estimated from point source measurements
501
CHAPTER 13
at one or more representative locations within the given scanner. This approach
typically provides satisfactory results within the central FOV of large, whole
body PET scanners. However, for PET systems with smaller ring diameters
(relative to the reconstruction FOV), such as animal systems, and for SPECT
systems with depth dependent resolution (and in particular with non-circular
orbits), it is desirable to use more accurate spatially variant resolution models.
The second category is using analytically calculated resolution functions
(usually spatially variant anisotropic kernels) for each location (LOR) as
determined based on analytical models of physical effects affecting the resolution.
This approach is usually limited to simple analytical models representing (or
approximating) only basic physical characteristics of the system. The resolution
kernels are usually calculated in real time during the reconstruction process when
they are needed within the forward and back projection calculations. In SPECT,
distance dependent collimator blurring requires convolution kernels that become
wider and, therefore, need more computation, with increasing distance to the
collimator. The computation time can be reduced considerably by integrating
an incremental blurring step into the projector (and back projector), based on
Gaussian diffusion. This method, developed by McCarthy and Miller in 1991, is
described in more detail in chapter22 of Ref.[13.5].
A more accurate but computationally very demanding approach is using
Monte Carlo simulations of the resolution functions based on a set of point sources
at various (ideally all) image locations. Setting up an accurate mathematical
model (transport equations tracing the photon paths through the detector system/
crystals) is relatively easy within the Monte Carlo simulations, compared to the
analytical approach of determining the resolution function. However, to obtain
sufficient statistics to get the desired accuracy of the shape of the resolution
functions is extremely time consuming. Consequently, simplifications often have
to be made in practice, such as determining the resolution kernels only at a set of
representative locations and interpolating/extrapolating from them the resolution
kernels at other locations.
The most accurate but also most involved approach is based on
experimental measurements of the system response by measuring physical point
sources at a set of image locations within the scanner. This is a tedious and very
time consuming process, involving point sources with long half-life isotopes and
usually requiring the use of accurate robotic stages to move the point source.
Among the biggest challenges is to accumulate a sufficient number of counts to
obtain an accurate point spread function, even at a limited number of locations.
Consequently, the actual resolution kernels used in the reconstruction model are
often estimated by fitting analytical functions (kernels) to the measured data,
rather than directly using the measured point spread functions.
502
At the conclusion of this subsection, it is worth making the following general

comment. In the light of the resolution modelling possibilities discussed above,
one might wonder whether it is worth spending energy and resources on building
new PET and SPECT systems with improved resolution properties. However,
although it has been shown in the literature that proper system models lead to
improved reconstructed image quality, they can never fully recover information
that has been lost through resolution effects and other instrumentation limitations.
Furthermore, due to the increased level of modelling, the system matrix becomes
more dense, and consequently the inverse problem (reconstruction) becomes
more ill-posed, thus making it impossible to attain perfect recovery for the
realistic data. There is no doubt that improved instrumentation as well as novel
and more accurate reconstruction models play an important role in improving
image quality and quantitative accuracy, and eventually increasing the general
clinical utility of emission tomography systems.
13.3.6.4. Motion corrections
Owing to the relatively long acquisition times, motion effects, caused by
patient movement and organ motion and deformation, cannot be avoided in
emission tomography. In the following, all of these effects are covered under the
simple term motion. With the continuous improvements of PET and SPECT
technology, leading to improved spatial resolution, signal to noise ratio, image
quality and accuracy of quantitative studies, corrections for motion effects
become more important. In fact, artefacts caused by motion are becoming
the single most important factor for image degradation, especially in PET
or PET/computed tomography (CT) imaging of the upper torso region. For
example, motion effects can lead to the loss of small lesions by blurring them out
completely in regions with strong motion (such as near the lower lung wall), or to
their misplacement into the wrong anatomical region (e.g. into the liver from the
lungs, or vice versa). Motion correction has become an important research topic;
however, a thorough discussion of this topic is out of the scope of this chapter
and interested readers are referred to the literature on this topic. In the following,
the main concepts of motion correction as dealt with within the reconstruction
process are outlined.
The two main sources of motion related artefacts in emission studies are
the motion during the emission scan and the discrepancy (caused by the motion)
between the attenuation and emission data. The motion during the emission scan
means that the emission paths (LORs) through the object (as considered in the
system matrix) change during the scan time. If this time dependent change is
not accounted for, the system model becomes inconsistent with the data, which
results in artefacts and motion blurring in the reconstructed images. On the other
503
CHAPTER 13
hand, the transmission scan (CT) is relatively short and can usually be done in
a breath-hold mode. Consequently, the attenuation image is usually motion-free
and captures only one particular patient position and organ configuration
(time frame). If the attenuation factors obtained from this fixed-time position
attenuation image are applied to the emission data acquired at different time
frames (or averaged over many time frames), this leads to artefacts in the
reconstructed images, which tend to be far more severe in PET than in SPECT.
This is, for example, most extremely pronounced at the bottom of the lungs
which can typically move several centimetres during the breathing cycle, causing
motion between two regions with very different attenuation coefficients.
Emission data motion: Correction approaches for motion during the
emission scan are discussed first. The first step is subdividing the data (in PET,
typically list-mode data) into a sufficient number of time frames to ensure that the
motion within each frame is small. For the organ movement, the frames can be
distributed over a period of the organ motion (e.g. breathing cycle). For the patient
motion, the frames would be typically longer and distributed throughout the
scan time. Knowledge about the motion can be obtained using external devices,
such as cameras with fiducial markers, expansion belts or breathing sensors for
respiratory motion, the electrocardiogram signal for cardiac motion, etc. There
are also a limited number of approaches for estimating the motion directly from
the data.
Once the data are subdivided into the set of the frames, the most
straightforward approach is to reconstruct data independently in each frame.
The problem with this approach is that the resulting images have a poor signal
to noise ratio because the acquired counts have been distributed into a number
of individual (now low count) frames. To improve the signal to noise ratio, the
reconstructed images for individual frames can be combined (averaged) after
they are registered (and properly deformed) to the reference time frame image.
However, for statistical non-linear iterative reconstruction algorithms, this is not
equivalent to (and typically of a lower quality than) the more elaborate motion
correction approaches, taking into account all of the acquired counts in a single
reconstruction, as discussed below.
For rigid motion (e.g. in brain imaging), the events on LORs (LORi) from
each time frame, or time position, can be corrected for motion by translation
(using affine transformations) into the new LORs (LORi) in the reference frame
(seeFig.13.21 (top right, solid line)), in which the events would be detected
if there were no motion. Reconstruction is then done in a single reference
frame using all acquired counts, leading to a better signal to noise ratio in the
reconstructed images. Care has to be taken with the detector normalization factors
so that the events are normalized using the proper factors (Ni) for the LORs on
which they were actually detected (and not into which they were translated).
504
Attenuation factors are obtained on the transformed lines (atti) through the
attenuation image in the reference frame. Care also has to be given to the proper
treatment of data LORs with events being translated into, or out of, the detector
gaps or detector ends. This is important, in particular for the calculation of the
sensitivity matrix, which then becomes a very time consuming process.
Frame k
Frame 0
Image estimate
Image estimate
Scanned object
Image estimate
Uncorrected image
Image estimate
FIG.
13.21. Illustration
of motion
corrections
for events
acquired
within line
of response
LORi with
FIG.13.21.
Illustration
of motion
corrections
for events
acquired
within
line of response
corresponding normalization Ni and attenuation atti factors. Left top: positions and shapes of the object in
LORi with corresponding normalization Ni and attenuation atti factors. Left top: positions
the reference time frame 0 and frame k. Left bottom: illustration of blurring in the reconstruction
and shapes
of the
in thewithout
reference
time
frame 0(attenuation
and frame factors
k. Left are
bottom:
illustration
combining
events
fromobject
all frames
motion
correction
also averaged
over
ofwhole
blurring
in of
thethereconstruction
events
from all within
framesthe
without
motion
Middle column:
processing
reference
time correction
frame. Right
the
range
frames atti0k).combining
line) of
hasthe
to be
transformed
the LORi
top:
LOR based factors
motion correction
frame k over
thethe
LORwhole
i (dashed
(attenuation
are also for
averaged
range
frames
atti0k).to Middle
(solid
line
for
rigid
motion,
dotted
line
for
non-rigid
motion)
which
represents
the
paths
that
the
photons
column: processing within the reference time frame. Right top: LOR based motion correction
would travel through the reference object if there were no motion. It should be noted that although the
has to
transformed
to the
LOR
line for
for frame
k the LOR
i (dashed line)
i (solid
LORs
are transformed,
the normalization
factors
arebeused
for the crystal
pairs
(LORs)
in which
the rigid
events
motion,
dotted
forthe
non-rigid
motion) factors
which are
represents
the paths that
while
used attenuation
for the transformed
pathsthe
(attphotons
bottom:
were
detected
(Ni),line
i). Right would
image
based
motion
the estimated
image
thethat
reference
frame
travel
through
thecorrection,
referenceincluding
object ifimage
theremorphing
were noofmotion.
It should
befrom
noted
although
(dashed
lines)are
intotransformed,
the given frame
line).
the LORs
the(solid
normalization
factors are used for the crystal pairs (LORs) in
while
attenuation
factors
are for the approach
transformed
which
were
detected into
(Ni), the
Oncethe
theevents
data are
subdivided
set the
of used
the frames,
the most
straightforward
is to
bottom: in
image
based The
motion
correction,
morphing
the
paths (attdata
reconstruct
independently
each frame.
problem
with thisincluding
approach image
is that the
resultingofimages
i). Right
have
a poor image
signal from
to noise
because
the (dashed
acquired lines)
countsinto
have
distributed
into line).
a number of
estimated
the ratio
reference
frame
thebeen
given
frame (solid
individual (now low count) frames. To improve the signal to noise ratio, the reconstructed images for
individual frames can be combined (averaged) after they are registered (and properly deformed) to the
referenceFor
time non-rigid
frame image.(elastic)
However,motion,
for statistical
non-linear
reconstruction
this is
which
is theiterative
case for
most of algorithms,
the practical
not equivalent to (and typically of a lower quality than) the more elaborate motion correction approaches,
applications,
the
motion
correction
procedures
become
quite
involved.
There
taking into account all of the acquired counts in a single reconstruction, as discussed below.
areFortwo
The first
approach
is to
derive
eachtransformations
time frame, or time
rigidbasic
motionpossibilities.
(e.g. in brain imaging),
the events
on LORs
(LOR
i) fromthe
position,
can be corrected
for events
motion by
translation
(using
affine
transformations)
into the new
LORs
of individual
paths of
(LORs)
from
each
frame
into the reference
frame
(LORi) in the reference frame (see Fig. 13.21 (top right, solid line)), in which the events would be detected
(topReconstruction
right, dotted
line)).
non-rigid
the
if (seeFig.13.21
there were no motion.
is then
doneFor
in athe
single
reference motion,
frame using
alltransformed
acquired counts,
leading
a better signal
noise ratio in
the reconstructed
Care has tolines
be taken
with the detector
pathstothrough
the toreference
object
frame areimages.
not straight
anymore,
thus
normalization factors so that the events are normalized using the proper factors (Ni) for the LORs on which
they were actually detected (and not into which they were translated). Attenuation factors are obtained on
the transformed lines (atti) through the attenuation image in the reference frame. Care also has to be given
505or
to the proper treatment of data LORs with events being translated into, or out of, the detector gaps
detector ends. This is important, in particular for the calculation of the sensitivity matrix, which then
becomes a very time consuming process.
For non-rigid (elastic) motion, which is the case for most of the practical applications, the motion
correction procedures become quite involved. There are two basic possibilities. The first approach is to
derive the transformations of individual paths of events (LORs) from each frame into the reference frame
CHAPTER 13
leading to very large computational demands for the calculations of the forward
and back projection operations. The same care for normalization, gaps and
detector ends has to be taken as above.
The second, more efficient, approach involves morphing the image
estimate (of the reference image) into the frame for which current events (LORs)
are being processed (seeFig.13.21 (bottom right, solid line)). It should be noted
that some pre-sorting of the data is considered, so that events from each frame
are processed together (using a common image morphing operation). Here, the
acquired LORs (LORi) and their normalization coefficients (Ni) are directly used
without modification. However, the sensitivity matrix still needs to be carefully
calculated, taking into consideration update and subset strategy, e.g. including
the morphing operation if subset data involve several frames. This is, however,
a simpler operation than in the LOR based case since the morphing is done in
the image domain. This image based approach is not only more efficient, but
also better reflects/models the actual data acquisition process during which the
acquired object is being changed (morphed).
Attenuation effects: In the following, it is considered that either attenuation
information for each time frame is available, for example, having a sequence of
CT scans for different time positions, or there is knowledge of the motion and
tools to morph a fixed-time position CT image to represent attenuation images at
individual time frames. It is further considered that tools are available to obtain
the motion transformation of data and/or images between the individual time
frames. If the emission data are stored or binned without any motion gating, they
represent motion-blurred emission information over the duration of the scan.
Using attenuation information for them for a fixed time position is not correct. It
would be better to pre-correct those data using proper attenuation factors for each
frame, but then the statistical properties (Poisson character) are lost due to the
pre-correction. A good compromise (although not theoretically exact) is to use
motion-blurred attenuation factors during the pre-correction or the reconstruction
process.
For data stored in multiple time frames, separate attenuation factors (or
their estimates) are used for each frame, such that they reflect attenuation factors
(for each LOR) at that particular time frame. For the case when there are multiple
CT images, this is simply obtained by calculation (forward projection) of the
attenuation coefficients for each frame from the representative CT image for that
frame. For the case when there is only one CT image, attenuation factors have
to be calculated on the modified LORs (for each time frame) in the LOR based
corrections, or to morph the attenuation image for each frame and then calculate
the attenuation factors from the morphed images in the image based corrections.
506
13.4. NOISE ESTIMATION

13.4.1. Noise propagation in filtered back projection
The pixel variance in an image reconstructed with FBP can be estimated
analytically, by propagating the uncorrelated Poisson noise in the data through
the reconstruction operation. The FBP algorithm can be written as:
( x, y) =
x
0
Y( x cos + y sin s)h( s) ds (13.86)
where h(s) is the convolution kernel, combining the inverse Fourier transform of
the ramp filter and a possible low-pass filter to suppress the noise.
The variance on the measured sinogram Y(s, ) data equals its expectation
Y (s, ); the covariance between two different sinogram values Y(s, ) and
Y(s, ) is zero. Consequently, the covariance between two reconstructed pixel
values (x,y) and (x, y) equals:
covar( ( x, y), ( x ', y')) =
x
0
Y( x cos + y sin s) ds
h(s)h( s + ( x ' x)cos + ( y' y)sin )
(13.87)
This integral is non-zero for almost all pairs of pixels. As h(s) is a high-pass
filter, neighbouring reconstruction pixels tend to have fairly strong negative
correlations. The correlation decreases with increasing distance between (x,y)
and (x, y). The variance is obtained by setting x=x and y=y, which produces:
var( ( x, y)) =
x
0
Y( x cos + y sin s) h( s) ds (13.88)
Figure13.22 shows the variance image of the FBP reconstruction of

a simulated PET sinogram of a heart phantom. The image was obtained by
reconstructing 400 sets of noisy PET data. The figure also shows a noise-free and
one of the noisy FBP images. The noise creates streaks that extend to the edge of
the image. As a result, the variance is non-zero in the entire image.
507
CHAPTER 13
True
With noise
Variance
(400 noise
realizations)
FIG.13.22. Simulated PET reconstructions of a heart phantom. Reconstructions were done

with filtered back projection (FBP), maximum-likelihood expectation-maximization (MLEM)
with Gaussian post-smoothing and with maximum a posteriori (MAP) using a quadratic prior.
For each algorithm, a noise-free and a noisy reconstruction are shown, and also the pixel
variance obtained from 400 independent Poisson noise realizations on the simulated PET
data. All reconstructions (first two rows) are shown on the same grey value scale. A second
scale was used to display the three variance images. The noisy FBP image contains negative
pixels (displayed in white with this scale).
13.4.2. Noise propagation in maximum-likelihood expectation-maximization

The noise analysis of MLEM (and MAP) reconstruction is more
complicated than that for FBP because these algorithms are non-linear. However,
the MLEM algorithm has some similarity with the WLS algorithm, which can
be described with matrix operations. The WLS reconstruction was described
previously; Eq.(13.45) is repeated here for convenience (the additive term was
assumed to be zero for simplicity):
= ( A'C y 1 A) 1 A'C y 1 y (13.89)
Cy is the covariance of the data, which is defined as C y = E( y y )( y y )' , where

E denotes the expectation and y is the expectation of y .
508
The covariance of the reconstruction is then:

C = E( )( )'
= ( A'C y 1 A) 1 A'C y 1E( y y )( y y )'C y 1 A( A'C y 1 A) 1 (13.90)
= ( A'C y 1 A) 1
This matrix gives the covariances between all possible pixel pairs in the
image produced by WLS reconstruction. The projection A and back projection
A have a low pass characteristic. Consequently, the inverse (ACy1A)1 acts as a
high-pass filter. It follows that neighbouring pixels of WLS reconstructions tend
to have strong negative correlations, as is the case with FBP. Owing to this, the
MLEM variance decreases rapidly with smoothing.
Figure13.22 shows mean and noisy reconstructions and variance images of
MLEM with Gaussian post-smoothing and MAP with a quadratic prior. For these
reconstructions, 16 iterations with 8 subsets were applied. MAP with a quadratic
prior produces fairly uniform variance, but with a position dependent resolution.
In contrast, post-smoothed MLEM produces fairly uniform spatial resolution, in
combination with a non-uniform variance.
REFERENCES
[13.1] LEWITT, R.M., MATEJ, S., Overview of methods for image reconstruction from
projections in emission computed tomography, Proc. IEEE Inst. Electr. Electron.
Eng. 91 (2003) 15881611.
[13.2] NATTERER, F., The Mathematics of Computerized Tomography, Society for
Industrial and Applied Mathematics (SIAM), Philadelphia, PA (1986).
[13.3] KAK, A.C., SLANEY, M., Principles of Computerized Tomographic Imaging, Society
for Industrial and Applied Mathematics (SIAM), Philadelphia, PA (1988).
[13.4] BARRETT, H.H., MYERS, K.J., Foundations of Image Science, John Wiley and
Sons, Hoboken, NJ (2004).
[13.5] WERNICK, M.N., AARSVOLD, J.N. (Eds), Emission Tomography, The
Fundamentals of PET and SPECT, Elsevier Academic Press (2004).
[13.6] NATTERER, F., Inversion of the attenuated radon transform, Inverse Probl. 17 (2001)
113119.
[13.7] XIA, W., LEWITT, R.M., EDHOLM, P.R., Fourier correction for spatially variant
collimator blurring in SPECT, IEEE Trans. Med. Imaging 14 (1995) 100115.
[13.8] DEFRISE, M., CLACK, R., TOWNSEND, D.W., Image reconstruction from
truncated, two-dimensional, parallel projections, Inverse Probl. 11 (1996) 287313.
509
CHAPTER 13
[13.9] DEFRISE, M., KUIJK, S., DECONINCK, F., A new three-dimensional reconstruction
method for positron cameras using plane detectors, Phys. Med. Biol. 33 (1988) 4351.
[13.10] KINAHAN, P.E., ROGERS, J.G., Analytic three-dimensional image reconstruction
using all detected events, IEEE Trans. Nucl. Sci. NS-36 (1990) 964968.
[13.11] DAUBE-WITHERSPOON, M.E., MUEHLLEHNER, G., Treatment of axial data in
three-dimensional PET, J. Nucl. Med. 28 (1987) 17171724.
[13.12] LEWITT, R.M., MUEHLLEHNER, G., KARP, J.S., Three-dimensional image
reconstruction for PET by multi-slice rebinning and axial image filtering, Phys. Med.
Biol. 39 (1994) 321339.
[13.13] DEFRISE, M., et al., Exact and approximate rebinning algorithms for 3D PET data,
IEEE Trans. Med. Imaging 16 (1997) 145158.
[13.14] DEFRISE, M., A factorization method for the 3D X-ray transform, Inverse Probl. 11
(1995) 983994.
[13.15] TOMITANI, T., Image reconstruction and noise evaluation in photon time-offlight assisted positron emission tomography, IEEE Trans. Nucl. Sci. NS-28 (1981)
45824589.
[13.16] QI, J., LEAHY, R.M., Iterative reconstruction techniques in emission computed
tomography, Phys. Med. Biol. 51 (2006) R541578.
[13.17] FESSLER, J.A., BOOTH, S.D., Conjugate-gradient preconditioning methods for shift
variant PET image reconstruction, IEEE Trans. Image Process. 8 (1999) 688699.
[13.18] PRESS, W.H., FLANNERY, B.P., TEUKOLSKY, S.A., VETTERLING, W.T.,
Numerical Recipes, The Art of Scientific Computing, Cambridge University Press
(1986).
[13.19] DE PIERRO, A.R., A modified expectation maximization algorithm for penalized
likelihood estimation in emission tomography, IEEE Trans. Med. Imaging 14 (1995)
132137.
[13.20] SHEPP, L.S., VARDI, Y., Maximum likelihood reconstruction for emission
tomography, IEEE Trans. Med. Imaging MI-1 (1982) 113122.
[13.21] DEMPSTER, A.P., LAIRD, N.M., RUBIN, D.B., Maximum likelihood from
incomplete data via the EM algorithm, J. R. Stat. Soc. Series B Stat. Methodol. 39
(1977) 138.
[13.22] PARRA, L., BARRETT, H.H., List-mode likelihood: EM algorithm and image
quality estimation demonstrated on 2D PET, IEEE Trans. Med. Imaging 17 2 (1998)
228235.
[13.23] READER, A.J., ERLANDSSON, K., FLOWER, M.A., OTT, R.J., Fast accurate
iterative reconstruction for low-statistics positron volume imaging, Phys. Med. Biol.
43 4 (1998) 835846.
[13.24] QI, J., Calculation of the sensitivity image in list-mode reconstruction, IEEE Trans.
Nucl. Sci. 53 (2006) 27462751.
510
[13.25] MATEJ, S., et al., Efficient 3D TOF PET reconstruction using view-grouped histo
images: DIRECT Direct Image Reconstruction for TOF, IEEE Trans. Med. Imaging
28 (2009) 739751.
[13.26] HUDSON, M.H., LARKIN, R.S., Accelerated image reconstruction using ordered
subsets of projection data, IEEE Trans. Med. Imaging 13 (1994) 601609.
[13.27] BROWNE, J., DE PIERRO, A.R., A row-action alternative to the EM algorithm
for maximizing likelihoods in emission tomography, IEEE Trans. Med. Imaging 15
(1996) 687699.
[13.28] DAUBE-WITHERSPOON, M.E., MATEJ, S., KARP, J.S., LEWITT, R.M.,
Application of the row action maximum likelihood algorithm with spherical basis
functions to clinical PET imaging, IEEE Trans. Nucl. Sci. 48 (2001) 2430.
[13.29] SNYDER, D.L., MILLER, M.I., THOMAS, L.J., Jr., POLITTE, D.G., Noise and
edge artefacts in maximum-likelihood reconstructions for emission tomography, IEEE
Trans. Med. Imaging MI-6 (1987) 228238.
[13.30] LEAHY, R.M., QI, J., Statistical approaches in quantitative positron emission
tomography, Stat. Comput. 10 (2000) 147165.
511
CHAPTER 14
NUCLEAR MEDICINE IMAGE DISPLAY
H. BERGMANN
Center for Medical Physics and Biomedical Engineering,
Medical University of Vienna,
Vienna, Austria
14.1. INTRODUCTION
The final step in a medical imaging procedure is to display the image(s)
on a suitable display system where it is presented to the medical specialist
for diagnostic interpretation. The display of hard copy images on X ray film
or photographic film has largely been replaced today by soft copy image
display systems with cathode ray tube (CRT) or liquid crystal display (LCD)
monitors as the image rendering device. Soft copy display requires a high
quality display monitor and a certain amount of image processing to optimize
the image both with respect to the properties of the display device and to some
psychophysiological properties of the human visual system. A soft copy display
system, therefore, consists of a display workstation providing some basic image
processing functions and the display monitor as the intrinsic display device.
Display devices of lower quality may be used during intermediate steps of the
acquisition and analysis of a patient study. Display monitors with a quality
suitable for diagnostic reading by the specialist medical doctor are called primary
devices, also known as diagnostic devices. Monitors with lower quality but good
enough to be used for positioning, processing of studies, presentation of images
in the wards, etc. are referred to as secondary devices or clinical devices.
Nuclear medicine images can be adequately displayed even for diagnostic
purposes on secondary devices. However, the increasing use of X ray images on
which to report jointly with images from nuclear medicine studies, such as those
generated by dual modality imaging, notably by positron emission tomography
(PET)/computed tomography (CT) and single photon emission computed
tomography (SPECT)/CT, requires display devices capable of visualizing
high resolution grey scale images at diagnostic quality, i.e. primary display
devices. Both grey scale and colour display devices are used, the latter playing
an important role in the display of processed nuclear medicine images and in
the display of overlaid images such as from registered dual modality imaging
studies.
512
Owing to the advances of picture archiving and communication systems

(PACSs), the location of a display device which used to be adjacent to the gamma
camera or PET scanner now widely varies, and can, for example, be in special
reporting rooms, patient wards and operating rooms. An important requirement
for display devices connected to a PACS is that, regardless of the display device
and the imaging modality used, the display of an image should be consistent in
appearance and presentation among the monitors and printers used at different
locations and under different environmental lighting conditions. A special
standard arising from this need was created specifying the requirements to
display grey scale images on different display devices [14.1]. For colour displays,
similarity in visual appearance is achieved by using an industry standard colour
management system (CMS) [14.2]. The requirements for inclusion of a CMS into
a PACS are addressed in Ref.[14.3].
The technology of soft copy display devices is presently experiencing
a rapid transition from CRT based monitors to LCD technologies, the latter
potentially offering better image quality, improved stability, reduced weight and
reduced costs.
Regardless of the technology, display hardware needs to operate at
maximum performance and needs to deliver consistent and reproducible results.
Quality assurance of a display device, both at the time of installation and at
regular intervals during its lifespan, ensures stability and optimum performance,
and should be considered an important component of the quality system in
nuclear medicine.
14.2. DIGITAL IMAGE DISPLAY AND VISUAL PERCEPTION
Emissive display devices such as CRTs or LCDs display a soft copy of
an image. Hard copies consist of transmissive film or of reflective prints, both
produced by printing devices.
The digital image to be displayed is represented as a rectangular matrix
of integer values. Each matrix element corresponds to a pixel of the image.
Regardless of the modality, the original intensity values as produced by the
acquisition device are integer values with a pixel depth of between 8 to 16 bits.
For display, the original images are stored in the memory of a computer, the
display workstation. The soft copy display system of a workstation consists of
the monitor which produces the visible image on the screen, and of the associated
display controller, also known as a graphics card or video card, which holds
the intensity values and carries out the conversion of the digital values to the
analogue signals which control the electronics of the display device.
513
CHAPTER 14
Originally, there is no colour in the image, but pseudo-colours are routinely

used in nuclear medicine images to improve the visibility or to emphasize special
features within an image. Another common use of colour is with overlay display
techniques of pairs of registered images, e.g. originating from dual modality
acquisitions such as PET/CT or SPECT/CT. Colour display devices, therefore,
play an increasingly important role in the visualization of medical images.
14.2.1. Display resolution
The resolution of a digital display device is commonly described in
terms of the number of distinct pixels in each dimension that can be displayed.
Displays are designed in such a way that individual pixels have an approximately
quadratic shape. The basic question arising is how many pixels are needed for
adequate visualization. The desirable lower limit of pixel size would be reached
if the human eye would not be able to distinguish between the individual pixels
which make up the screen. Assuming a spatial resolution of the human eye of
1 arcminute and an average reading distance for an X ray image of 65cm, a
human observer would not be able to discern two adjacent pixels as being
different for pixels smaller than about 0.18mm. A modern 3 megapixel (MP)
colour LCD monitor with a screen diagonal of 540mm (~21 in) has pixel
matrix dimensions of 15362048, resulting in a pixel size of 0.21mm which
is close to the resolution limit of the eye, so that individual pixels are almost
indistinguishable. Monitors of this quality are already used routinely in radiology
as primary display devices except for mammography. A 5 MP LCD monitor of
the same screen size would have pixel matrix dimensions of 20482560 and
linear pixel dimensions of 0.17mm. Monitors with such high resolution are
accepted as primary devices for reading digital mammography images. Limits of
display resolution can be overcome by magnification and interpolation techniques
included as standard features in the display workstation.
LCD monitors are composed of a large number of liquid crystals. This
number represents the native resolution of the display device since each pixel
can be addressed individually to change brightness and colour. The display of
an image is best if each pixel of the image maps to a pixel of the screen. If the
mapping requires interpolation of the image pixels to the screen pixels, the image
loses sharpness. A CRT display, in contrast, can change screen pixel size without
loss of image sharpness by changing deflection and modulation frequencies.
Several display resolution values can, therefore, be used with equal sharpness.
A display issue specific to nuclear medicine arises from the fact that the
matrix dimensions of the original acquired or reconstructed image are much
smaller than the display resolution would permit to display. The small original
image size is due to both the poor spatial resolution of a nuclear medicine
514
imaging device and its noise characteristics. As a rule of thumb, the sampling
size used for a nuclear medicine image preserves the information content in the
image when it is approximately one third of the full width at half maximum
(FWHM) of the spatial resolution of the acquisition system. Thus, a scintillation
camera equipped with, for example, a high resolution collimator, a system spatial
resolution of 8mm FWHM and a field of view of 540mm400mm would
require a matrix size of at most 256256 pixels to preserve the information
content transmitted by the camera. Even a current commercial off the shelf
(COTS) display device, on the other hand, hasminimum pixel dimensions of
10241280. Displaying the original image matrix at native pixel resolution
would result in an image too small for visual interpretation. It is, therefore,
essential for the image to be magnified to occupy a reasonable sized part of the
available screen area. Straightforward magnification using a simple interpolation
such as the nearest neighbour interpolation scheme would result in a block
structure with clearly visible square elements which strongly interfere with the
intensity changes created by the true structure of the object generating artefacts
in the interpretation. Magnification is necessary and can be done without artefact
generation by a suitable interpolation algorithm that generates smooth transitions
between screen pixels while preserving the intensity variations within the original
image. It is the task of the display workstation to provide software for visualizing
this type of image.
14.2.2. Contrast resolution
This refers to the number of intensity levels which an observer can perceive
for a given display. It is referred to as perceived dynamic range (PDR).
Brightness refers to the emitted luminance on screen and is measured in
candelas per square metre (cd/m2). The maximum brightness of a monitor is
an important quality parameter. Specifications of medical display devices also
include the calibrated maximum brightness which is lower but is recommended
for primary devices to ensure that the maximum luminance can be kept constant
during the lifespan of the display device. Typical values for a primary device LCD
monitor are 700 and 500 cd/m2 for the maximum and the calibrated maximum
luminance, respectively.
The dynamic range of a display monitor is defined as the ratio between the
highest and the lowest luminance (brightness) a monitor is capable of displaying.
The dynamic range is highest if measured in the absence of ambient light. It is
then called contrast ratio (CR=LH/LL) and is the figure usually quoted by vendors
in the specifications. A typical CR of a grey scale primary LCD monitor is 700:1,
measured in a dark reading room. If luminance values are measured with ambient
light present, which is the scenario in practice, CR is replaced by the luminance
515
CHAPTER 14
ratio (LR=LH/LL), which is the ratio of the highest and the lowest luminance
values including the effect of ambient light. It can be considerably smaller than
the CR, since the effect of ambient lighting is added as a luminance Lamb to both
theminimum and the maximum luminances. The CR is related to the PDR, but
its potential usefulness as a predictor of monitor performance suffers from the
lack of standardized measurement procedures and the effect of ambient light. The
dark room CR is a common performance parameter quoted by manufacturers.
The PDR is the number of intensity levels an observer can actually
distinguish on a display. It can be estimated based on the concept of just
noticeable differences (JNDs). The JND is the luminance difference of a given
target under given viewing conditions that the average human observer can just
perceive. The measured JND depends strongly on the conditions under which
the experiment is performed, for example, on the size, shape and position of the
target. The PDR is defined as the number of JNDs across the dynamic range of
a display. The PDR for grey scale displays has been assessed in Ref.[14.4] to
be around a hundred. The number of intensity values which a pixel of the digital
image can hold is much higher. The pixel of an image matrix is usually 1 or
2 bytes deep. It is an integer number between 256 and 65536, and is given by
the pixel depth of the image matrix. It is a further task of the display system to
scale the original intensity values to a range compatible with the performance
of the human observer. It is common to use 256 intensity values to control the
brightness of a display device as this is sufficient to produce a sequence of
brightness levels perceived as continuous by the human observer.
Colour displays using pseudo-colour scales can extend the PDR to
about 1.5 times that of a grey scale display. This has been demonstrated for a
heated-object scale which has the additional advantage of producing a natural
image [14.4]. Owing to the enormous number of possible colour scales and the
fact that the majority of them produce unnatural images, the concept of JNDs,
while valid in principle, cannot be transferred directly to colour displays.
14.3. DISPLAY DEVICE HARDWARE
14.3.1. Display controller
The display controller, also known as a video card or graphics card, is the
interface between the computer and the display device. Its main components are
the graphical processing unit (GPU), the video BIOS, the video memory and the
random access memory digital to analogue converter (RAMDAC). The GPU
is a fast, specialized processor optimized for graphics and image processing
operations. The video memory holds the data to be displayed. The capacity of
516
current COTS graphics cards ranges from 128 MB to 4 GB which is sufficient

to store even a large sequence of images. This permits the rapid change between
images on the screen for, for example, cine display, rapid browsing through a
sequence of tomographic slices or the use of overlay techniques to display
simultaneously additional information such as text, markers or regions of interest
without having to modify the image data. The output signals generated by the
RAMDAC depend on the monitor type. For CRT monitors, the RAMDAC
generates analogue positioning and intensity signals and additional control
signals for the deflection and synchronization of the electron beam. The output
of the graphics card is via a video graphics array connector. For current LCD
displays, the graphics card provides standardized digital output signals via the
digital visual interface connector which avoids image distortion and electrical
noise, and can be configured to use the native spatial resolution of the LCD
display directly.
Lookup tables (LUTs) constitute a crucial part of the video memory because
they play an important role in the implementation of intensity transformations
and in the display of colours. An LUT contains the digital values which are
converted by the RAMDAC to the analogue intensity values driving the monitor.
These values are called digital driving levels (DDLs). The range of DDLs is
typically 8 bit, i.e. from 0 to 255. For medical displays, the range can be larger,
up to 12 bit. The maximum value contained in an LUT produces the maximum
brightness the screen can display. Since there is some latitude in the maximum
brightness a screen can display, it is adjusted by the monitor hardware controls
or firmware, and in practice follows manufacturers recommendations to ensure
optimum performance with respect to both image quality and life expectancy of
the monitor. The luminance values generated by the sequence of available DDLs
(e.g. 0255) produce the characteristic curve of the display device.
The intensity values of an image stored in video memory are mapped to
the values in an LUT. The mapping transformation associates each intensity with
an LUT index. In the case of a colour display, a triple of DDLs, one each for the
three primary colours red, green and blue, is used for a pixel. The LUT consisting
of triples of primary DDL values is referred to as a colour lookup table (CLUT).
For a medical image, which by nature is a grey scale image, a pseudo-colour
image is generated by a mapping transformation which associates the intensity
value of an image pixel to an appropriate index of a CLUT.
14.3.2. Cathode ray tube
The CRT is a vacuum tube containing an electron gun and a fluorescent
screen (Fig.14.1). The electron gun produces electrons which are focused by
a system of magnetic and electrostatic lenses into a pencil beam. The electron
517
CHAPTER 14
beam is accelerated by a positive high voltage applied to the anode towards

the fluorescent screen. The screen is covered with a crystalline phosphorescent
coating that produces a visible light spot when hit by the electron beam. The
light distribution of the spot is given by a 2D Gaussian function. The directional
dependence of the intensity of the emitted light follows Lamberts cosine law,
implying that the apparent luminance of the screen is independent of the viewing
angle. Using the deflecting coils attached to the collar of the tube, the beam
is made to scan the screen area in a rectangular pattern. At the same time, the
intensity of the electron beam is controlled by the control grid, thereby producing
different light intensities. The digital image matrix containing the numerical
intensity values is visualized by synchronizing scanning motion and intensity
modulation given by the LUT to produce an intensity pattern, the visual image,
on the screen. The elements of the matrix occupy a rectangular grid on the screen,
with the luminance of the centre of each grid point corresponding to the LUT
value of the corresponding matrix element.
FIG.14.1. Diagram of a cathode ray tube.
Colour CRTs use three different phosphors which emit red, green and blue
light, respectively. The phosphors are packed together in clusters called triads
or in stripes. Colour CRTs have three electron guns, one for each primary colour.
Each guns beam reaches the dots of exactly one type of phosphor. A mask close
to the screen absorbs electrons that would otherwise hit the wrong phosphor. The
triads or stripes are so small that the intensities of the primary colours merge in
the eye to produce the desired colour.
518
14.3.3. Liquid crystal display panel

An LCD display panel consists of a rectangular array of liquid crystal
cells in front of a light source, the backlight. Each cell acts as a tiny light valve
which transmits the backlight to an extent determined by a voltage applied to the
liquid crystal. The image on the screen is formed by applying voltages to each
cell separately, thereby modulating the light intensity into the desired intensity
pattern.
A typical liquid crystal cell (Fig.14.2) consists of a liquid crystal in twisted
nematic phase between two glass plates, G, coated with alignment layers (not
shown) that precisely twist the liquid crystal by 90 when no external field is
present (left diagram). Light from the backside is polarized by polarizer P1 and
rotated by the crystal structure. The second polarizer P2, set at 90 to P1, then
permits the light to pass. If a voltage is applied to the two transparent electrodes,
E1 and E2, the nematics realign themselves (right diagram) and the polarized
light is blocked by P2. Partial realignment is achieved by varying the voltage and
permits the transmitted intensity to vary.
FIG.14.2. Illustration of construction and operation of a single pixel of a twisted nematic

liquid crystal cell. No voltage applied=OFF state (left diagram); voltage applied=ON state
(right diagram) (courtesy of M. Schadt).
In colour LCDs, each individual pixel is divided into three cells, or

subpixels, which are coloured red, green and blue, respectively, by additional
filters. Each subpixel can be controlled independently, so that thousands or
millions of possible colours can be obtained for each pixel.
519
CHAPTER 14
An active matrix LCD is the predominant type of flat panel display. It is

the standard display device used as general computer displays, in notebooks and
increasingly as high quality displays for medical imaging. The active matrix
design permits switching each pixel individually by applying a row and column
addressing scheme. It is implemented by thin film transistor technology which
supplies each pixel of the display with its own transistor. The circuit is made by
depositing a thin film of silicon on the glass surface where the transistors are
fabricated. The transistors take up only a small fraction of the surface and the rest
of the silicon film is etched away to let the light pass through (Fig.14.3).
LCD displays using twisted nematic liquid crystals exhibit a strong
dependence of display brightness and colour on viewing angle. Developments
in technology have considerably improved the angular viewing response. The
preferred technique used for medical display devices is currently the in-plane
switching (IPS) technology. IPS aligns the crystals horizontally and applies
the voltage to realign the liquid crystal structure to both ends of the cell.
Non-uniformity of both luminance and luminance ratio of an LCD is expressed
as a function of viewing angle (from the normal to the display surface) for
horizontal and vertical directions separately.
Fig.14.3. A diagram of the pixel layout. Each liquid crystal pixel is connected to a transistor
which provides the voltage that controls the brightness. The pixel is addressed by a rowcolumn
scheme.
520
14.3.4. Hard copy devices

Although reporting is increasingly performed using soft copy displays and
PACS facilities, there is still a need for hard copies of images, such as for use
in the operating room or to be sent to referring physicians. Hard copy images
of diagnostic quality are printed on X ray film and use high resolution laser
printing and dry processing. In nuclear medicine, where the original images of a
diagnostic procedure often use pseudo-colours for the final display of the results,
even cheap mainstream colour printers may adequately be used. The long term
storage properties of hard copy media are not an issue when used in connection
with a PACS, where images are stored in electronic form. Therefore, technologies
such as dry laser film, thermal printers, colour laser printers or inkjet printers are
all acceptable hardcopy output devices.
The spatial resolution of a printer is given in dots per inch (dpi), a measure
of spatial printing density. It is defined as the number of individual dots that can
be placed within the span of 1 in (2.54cm).
14.3.4.1. Film laser imager
Dry laser imagers print X ray images on transparent film with the same
quality as available with conventional X ray film. Spatial resolution is up
to 650 dpi, adequate for diagnostic quality output for all imaging procedures,
including mammography. Contrast resolution depends on film quality and can
reach a Dmax of up to 4.0.
14.3.4.2. Colour printer
Mainstream colour laser printers produce cheap grey scale and colour
output of images with a spatial resolution of typically 6001200 dpi. For normal
paper, the CR is low. Image quality can be improved by using special paper with
a smooth surface for improved toner gloss and sharpness.
Inkjet printers can have a spatial resolution of up to 96002400 dpi. With
special photo paper, excellent image quality equivalent to colour photographs can
be achieved. The stability of the printout is known to be somewhat fragile, with
the image fading within a couple of years even under optimal storage conditions.
14.4. GREY SCALE DISPLAY
Nuclear medicine images do not require the same high quality grey scale
displays as is necessary for the display of X ray images. This can be attributed to
521
CHAPTER 14
the fact that the nuclear medicine image is a low count image with considerable
statistical fluctuations, making the comparison of tiny intensity differences
meaningless. A main difference to diagnostic X ray reporting is the fact that
colour in the image was recognized early as a helpful technique to improve
diagnostic reading and a tradition of visualization in colour has been established.
Therefore, images and analysis of results, in particular curves and functional or
metabolic images, are preferably displayed using colours. The display is usually
done on workstations with special nuclear medicine software and using current
COTS colour LCD screens as standard display devices. Typical screen sizes
are from 20 to 24 in and native display resolutions from 10241280 pixels
to 12001600 pixels. Depending on the capabilities of the nuclear medicine
display workstations software, several monitors can be used simultaneously.
The need to be able to perform concurrent diagnostic reporting on X ray
images, generated by dual mode acquisition techniques such as PET/CT and
SPECT/CT, and the inclusion of images from other modalities via PACS in the
reporting session, requires the use of grey scale display devices of diagnostic
quality (primary devices) at the nuclear medicine display workstation.
Both CRT and LCD display devices are available with spatial resolution
and CRs satisfying the requirements for a primary device. LCD displays are
rapidly replacing CRT displays for several reasons:
LCDs typically have about twice the brightness of CRTs. An overall
brighter image is less sensitive to changes in the level of ambient light and
is preferred for reporting.
LCD monitors exhibit no geometric distortion.
LCDs have a weight that is about one third of that of a comparable CRT.
LCDs are less prone to detrimental ageing effects.
LCDs are less expensive.
Furthermore, high quality colour LCD devices can be used as grey scale
primary devices, which is not feasible for a colour CRT monitor.
14.4.1. Grey scale standard display function
Todays ubiquity of PACSs enables deployment of display devices at all
locations where access to medical images is needed. The main challenge when
using different display devices in a PACS is to ensure that an image presented to
an observer appears identical irrespective of the display device used, be it a CRT
based or LCD based soft copy display, or hard copy displays, such as film laser
printers or paper printers. The Digital Imaging and Communications in Medicine
(DICOM) grey scale standard display function (GSDF) offers a strategy that
522
ensures that a medical image displayed or printed at any workstation or printing

system for which the GSDF is implemented has the same visual appearance,
within the capabilities of the particular display device [14.1]. This implies that
a display device complying with the GSDF must be standardized and calibrated,
and that a scheme of regular quality control is required for the display systems on
the PACS. Colour display systems may also be used for the purpose of displaying
grey scale images if calibrated to the GSDF [14.3].
Luminance (cd/m2)
DICOM grey scale standard display function
Just noticeable difference index
FIG.14.4. Digital Imaging and Communications in Medicine (DICOM) grey scale standard
display function.
The visual appearance of a native image as produced by the acquisition

device (gamma camera, PET scanner, CT scanner) depends, if no corrections
are applied, on the characteristic curve of the particular display device used at a
display workstation. An image displayed using the characteristic curves inherent
to a particular device could be significantly different in visual perception from
the optimal presentation. Part 14 of the DICOM standard [14.1] standardizes
the display of grey scale images. It does so by introducing the GSDF which can
be seen as a universal characteristic curve (Fig.14.4). The GSDF is based on
human contrast sensitivity. It covers a luminance range from 0.05 to 4000 cd/m2.
Theminimum luminance is the lowest that can be used in practice with a CRT
display, whereas the maximum luminance is slightly above the luminance of a
very bright unattenuated light box used for the examination of mammography
X ray films, so that it covers the range of luminance values of all display
523
CHAPTER 14
devices in current use. Human contrast sensitivity is non-linear within this

range. Perceptual similarity of a displayed image is achieved by linearizing the
GSDF with respect to contrast sensitivity. This is done by introducing the JND
index. One step in the JND index corresponds to a luminance difference that is
just noticeable, regardless of the mean luminance level. The DICOM standard
contains the standard GSDF as a tabulation of luminance (brightness) against
the JND index. Table14.1 shows the first and the last few JND indices of the
tabulation. It can be seen clearly that the relative changes of luminance need to
be much larger on the dark side of the curve than on the bright side to achieve a
JND.
TABLE14.1. TABULATED JUST NOTICEABLE DIFFERENCE INDICES
OF THE Grey scale standard display function
Just noticeable difference
Luminance (cd/m2)
0.0500
0.0547
0.0594
0.0643
1021
3941.8580
1022
3967.5470
1023
3993.4040
Note: The relative difference between the luminance of consecutive just noticeable difference
indices is much higher for low indices (~9%) than for high indices (~0.6%).
An individual display device with a luminance range LminLmax and a

DDL range of, for example, 8 bits exhibits a characteristic luminance curve as
a function of the DDL as shown in Fig.14.5. The device specific characteristic
curve will usually not match the corresponding segment of the GSDF. A
transformation is needed, which is implemented as an LUT. The LUT will map
a DDL Ds which should produce the standard luminance value Ls to the value
Dm, so that for input level Ds the transformed value Dm will produce the correct
luminance as required by the GSDF. The transformation LUT correcting for the
524
deviations of the characteristic curve of a specific display system from the GSDF
may be implemented directly in the display device or in the video memory of the
display controller. The result of the transformation is that the modified DDLs
operating the display will generate a characteristic curve identical to the GSDF.
FIG.14.5. Mapping of digital driving level Ds to the value Dm, so that for input level Ds which
should produce the standard luminance value Ls the transformed value Dm will produce the
correct luminance as given by the grey scale standard display function.
14.5. COLOUR DISPLAY

The human eye can distinguish millions of different colours. The full range
of colours the average human can see is given by the spectrum of sunlight. Each
colour can be characterized by three coordinates representing the colour as a
mixture of three primary colours in a colour space.
One of the first colour spaces introduced in 1931 is the International
Commission on Illumination (CIE) xyz colour space [14.5]. It is derived from a
model of human colour perception and uses three tri-stimulus values to compose
a colour. It is designed in such a way that one of the three coordinates defines
luminance (brightness); the other two coordinates represent the chromaticity
(hue). This leads to the well known CIE 1931 chromaticity diagram with its
typical horseshoe shape in which all colours the human visual system can perceive
are represented as a function of two coordinates, x and y (Fig.14.6). The third
525
CHAPtER 14
coordinate, representing brightness, would only change the saturation of a colour,

so, for example, varying the brightness coordinate for the chromaticity white
would run through all levels of grey from black to the maximum white a display
device can render.
FIG. 14.6. International Commission on Illumination xy chromaticity diagram. The outer

curved boundary is the spectral locus, with wavelengths shown in nanometres.
another frequently used colour space is the red, green, blue (rGb) space, a
natural colour space for a crT or lcd colour monitor. it uses as coordinates the
intensities of the red, green and blue primary colours to generate a colour pixel
(fig. 14.7).
The colour space used for hard copy printers is the cyan, magenta, yellow,
key (black) (cMyk) space.
The quality of the colour image depends on the colour depth (the range of
colour intensities) with which each subpixel contributes. colour quality increases
with subpixel depth. a common classification of the display controllers ability
to reproduce colours is: 8 bit colour (can display 256 colours), 15/16 bit colour
(high colour: can display 65 536 colours), 24 bit colour (true colour: can display
16 777 216 colours) and 30/36/48 bit colour (deep colour: can typically display
over a billion colours).
526
nUCLEAR MEDICInE IMAGE DIsPLAY
Gray line
FIG. 14.7. Red, green, blue colour cube with a grey line as diagonal. The number of (r, g, b)
voxels, i.e. colours available, depends on the bit depth of each coordinate. A depth of 8 bits for
each component would result in 16 777 216 colours.
a nuclear medicine display controller can usually handle true colour pixel
depths, with 8 bits available for each primary colour.
colour was utilized in nuclear medicine already at an early stage of the
development of digital displays. since the original image data contain no colour
information, the allocation of colour to an image pixel can be freely chosen. The
allocation takes the form of a cluT. conceptually, the cluT is an array structure
containing the colour coordinates for each colour included in the table. a colour
is defined by three values representing the intensities of the red, green and blue
subpixels. each pixel intensity in the image maps to an array index of the luT, so
that each intensity is associated with a particular colour. This is accomplished by
a transformation algorithm. The transformation is usually carried out by the GPu
of the display controller. The cluT is stored in the memory of the graphics card.
The luT is usually much smaller in size than the image. usual cluTs contain
64256 elements, respectively colours. an advantage of a cluT is that colours
can be changed by changing the luT, resulting in better display performance.
it is worthwhile noting that for a real world colour image, the colour of each
pixel is determined by the image itself and cannot be arbitrarily associated with
a colour such as is the case for pseudo-colour display. Thus, the quality of a real
world image increases the larger the number of colours that can be reproduced.
using a cluT for a colour image of the real world implies a loss of quality, as
527
CHAPTER 14
can easily be seen on images on the Internet which use CLUTs with typically
64 colours to save on image size. The addition of colour information to native
nuclear medicine and X ray images always results in a pseudo-colour image, with
the colours chosen by the user.
A modern nuclear medicine system typically uses 1632 different
CLUTs. The choice of colours is a complex issue. A continuous colour scale
can be achieved if the individual components vary slowly and continuously.
Pseudo-colour can be used to increase the PDR relative to grey scale; other
CLUTs may emphasize regions with a specific intensity as, for example, in
the case when performing a Fourier analysis of the beating heart to highlight
amplitude and phase information.
14.5.1. Colour and colour gamut
As with grey scale images, it is expected that a colour image displayed on
a PACS display device has the same colour appearance regardless of the type or
the individual characteristics of the display device. Fortunately, the problem of
producing digital colour images with the same perception of colours regardless
of the display device, including display monitors and hard copy printers, has
already been resolved by the printing industry and the photographic industry.
Since each colour is a unique entity, it is to be expected that unambiguous
transformations exist between the coordinates representing the colour in different
colour spaces. Such transformations are indeed available and are the basis of a
CMS. The purpose of a CMS is to produce a colour image that is perceived as
being the same by a human observer regardless of the output device used.
The gamut or colour gamut is defined as the entire range of colours
a particular display device can reproduce. The gamut depends on the type of
display and on design characteristics. The number of vertices of the gamut is
given by the number of primary colours used to compose a colour. In the case
of an LCD or a CTR monitor, the three primary colours, red, green and blue,
are used to produce a colour. For a printer, the colours of several inks or dyes
can be mixed to produce a colour on paper. Most printers can create dots in
a total of six colours which are cyan, yellow, magenta, red (which combines
yellow and magenta), green (yellow plus cyan) and blue (cyan plus magenta).
Typical gamuts for an LCD monitor and for a printer are shown in Fig.14.8. It
is obvious that the monitor can display colours unavailable to the printer and
vice versa. The International Color Consortium (ICC) has published procedures
including colour transformations (CMS) that ensure that a colour image that is
displayed on, for example, a monitor has the same appearance on, for example,
a colour printout [14.6]. The system is based on describing the colour properties
of a colour display device by an ICC colour profile. The colour profile contains
528
manufacturer provided or, preferably, the measured gamut of the individual

display device in a format which permits transformation of the colours between
the representation on the display device and an intermediary colour space,
for which the cie xyz space or the cie lab space are used. The intermediary
colour space acts as a colour reference and is called the profile connection space
(Pcs). The Pcs is utilized by dicoM [14.3], analogous to the Gsdf for grey
scale displays, as a reference space for the transformation of colours from one
colour display device to another. unlike the Gsdf, it does not, however, claim
to linearize perceived contrast. The colours of an individual display device can
be transformed with the help of the icc profiles to any other display device,
including hard copy colour printers while maintaining the same visual perception
of the colours. for colours available on one device but not on the other, the Pcs
substitutes colours similar in perception to the missing ones. in order to make
the Pcs work, it is necessary that all display systems involved have their icc
profiles available. The dicoM standard formalizes the information required for
colour management by adding the necessary tags to the data dictionary, so that
in a medical Pacs the colour transformations are carried out by the cMs in a
manner transparent to the user.
FIG. 14.8. Typical gamuts for a liquid crystal display monitor and a colour printer. The large
non-overlapping areas of the colours which cannot be reproduced by the other device and
must be substituted by similar colours should be noted.
529
CHAPtER 14
14.6. iMaGe disPlay MaNiPulaTioN

14.6.1. Histograms
The intensity histogram of an image represents the distribution of the
grey values in an image. it is obtained by dividing the range of grey values into
intervals of equal width, the bins, and calculating the number of pixels with
intensity values falling into each bin. The number of bins to store the frequencies
can be freely chosen, but the most informative displays are obtained with bin
numbers between 32 and 256. The graphical display of the histogram transmits a
rough idea about the distribution of intensities (fig. 14.9).
FIG. 14.9. Transverse CT slice at the height of the heart (left), with the corresponding intensity
histogram, using only the pixels within the region surrounding the trunk. The number of bins
is 256. Even when excluding all background pixels, the unequal distribution of intensities
is seen, especially the lack of high intensity values which is in agreement with the small
proportion of bony structure in the image.
14.6.2. Windowing and thresholding

The most basic intensity transformations used in image display are to
transform a pixel intensity I to a grey scale intensity value r within the available
grey scale range Q of the display monitor:
r = T(I)
(14.1)
Q is normally in the range 0...255. The transformations do not take into account
the values of surrounding pixels; each pixel is processed independently.
530
Windowing and thresholding are linear intensity transformations of that

type. They provide an easy method to emphasize contrast and visibility of
structures in areas of interest by only mapping intensity values within an intensity
window defined by a threshold and a window width to the available range of
brightness values. Values below the threshold are set to black; values above the
upper level are set to maximum display intensity. Thus, for an intensity threshold
level t and a window width w, the pixel intensity I is transformed into the grey
scale value r according to:
I , t I t + w
r = 0, I < t
Q, I > t + w
(14.2)
with = Q/W.
Windowing and thresholding may be hardware implemented, i.e. the
values may be changed by turning knobs on the monitor or the console or, more
frequently, by software implementation using mouse movements, sliders or the
arrows on the keyboard of the display workstation. The diagnostic value offered
Frequency
Bone window
Mediastinal window
Lung window
Hounsfield units
FIG. 14.10. CT slice from Fig. 14.9 with typical lung and mediastinal windowing (upper row
from left to right), a bone window (bottom left) and a histogram with corresponding linear
window functions (bottom right).
531
CHAPTER 14
by windowing and thresholding can be appreciated when using typical CT

windows for bone, mediastinum and lung. This is shown for a transaxial CT slice
through the chest, together with the range of intensity values actually visualized
out of the total histogram (Fig.14.10).
Simple algorithms may be successfully used to perform automatic
windowing and thresholding using histogram data, such asminimum and
maximum intensity or setting the threshold and window width in such a way
that a small percentage of the lowermost and uppermost intensity values are
discarded. Suitable values are between 0.5 and 2%.
14.6.3. Histogram equalization
Image intensity values may utilize the range of display intensities
inefficiently. The CT slice of Fig.14.9 is a typical example of a medical image
and demonstrates that most of the intensity values are the Hounsfield units for
soft tissue and the lung.
Histogram equalization aims at utilizing each grey scale level available
for display with the same frequency. If all intensity values were present in
equal numbers in the image, the histogram would be flat, and the corresponding
cumulative density function would increase linearly. A redistribution of intensity
values s to approximately equally distributed grey scale intensity values r can be
achieved using the transformation:
req=CDF(I)(Q 1)/(MN)
(14.3)
where
CDF(I) is the cumulative density function of the original image;
Q
is the available range of grey scale values;
and the image size is MN pixels.
For more details, see Ref.[14.7]. Figure14.11 demonstrates the effect of
histogram equalization using the standard algorithm of the image processing
software package ImageJ [14.8] on the CT slice of Fig.14.9. In the processed
image, the structures of both the bronchi of the lung and the ribs are visualized
in the same image without underflow or overflow and with approximately the
same information content as in the three windowed images of Fig.14.10 together.
The drawbacks of the method are that the visual appearance of an image depends
on the shape of the histogram and may, therefore, be significantly different
between patients, and the fact that the resulting intensity data can no longer be
532
used to extract quantitative information. The latter is nicely shown by the range
of intensity values in the histogram of fig. 14.11, which no longer shows the
familiar range of cT numbers.
FIG. 14.11. CT slice from Fig. 14.9 after histogram equalization with a corresponding intensity
histogram, using only the pixels within the region surrounding the trunk. The number of bins
is 256. The cumulative density function is now approximately linear. The intensity values are
no longer related to Hounsfield units. Owing to the better distribution of the intensity values,
all of the structures of interest, including the bone and the bronchi of the lung, are visualized
simultaneously.
14.7. VisualiZaTioN of VoluMe daTa

14.7.1. slice mode
an image volume dataset consists of a series of adjacent image slices
through the patients body. The slices can be displayed sequentially with manual
stepping through the images or automatically as a movie, or they can be displayed
simultaneously as a montage of several images. specialized viewing software
offers easy ways to manipulate the display further and permits, for example,
zooming and panning. Panning consists of quickly moving around a zoomed
image too large to be displayed completely on the screen by utilizing the mouse,
a joystick or a scroll wheel.
from the original slices, orthogonal views can easily be calculated by
rearranging the pixel matrix. Presenting the orthogonal views simultaneously on
the screen facilitates the anatomical location of structures. slices with oblique
orientations can also be calculated. in myocardial sPecT and PeT, reorientation
along the long and the short axes of the left ventricle are the standard display for
533
CHAPtER 14
visualization (fig. 14.12). a substantial gain in anatomical information can be

achieved by using linked cursors. The technique consists of moving a cursor
in one image while a second cursor is simultaneously moved by software to
identical locations in the other views.
FIG. 14.12. Orthogonal views of myocardial perfusion SPECT with orientation of the slices
along the long axis of the heart. The upper row shows an original transaxial slice through
the myocardium with the white line indicating the long axis (left) and a sagittal slice (right).
The bottom row shows the reoriented views with the vertical and horizontal slices through
the long axis, and a slice perpendicular to the long axis (from left to right). (Courtesy of
B. Knig, Hanuschkrankenhaus, Vienna.)
14.7.2. Volume mode

Volume mode display refers to techniques which extract the information
about structures in the 3-d image dataset by selecting intensity information
directly from the volume dataset and projecting the selected values on the display
screen. The ray casting technique projects a line from a viewpoint through
the data starting from a pixel on the display screen. it calculates the value of
interest using the image intensities along its path (fig. 14.13). less frequently
used in nuclear medicine are splatting techniques consisting of the projection
534
of possibly modified image voxels on the display screen; these techniques will
not be considered further here. Details can be found, for example, in Ref.[14.9].
The dominant ray casting geometry in nuclear medicine applications and in dual
mode imaging is parallel projection. Perspective projection is predominantly
used in virtual endoscopy and is not yet used routinely in dual mode imaging.
14.7.2.1. Transmission type volume rendering
Maximum intensity projection (MIP) consists of projecting the maximum
intensity value encountered along the trajectory of the ray through the data volume
on the corresponding screen pixel. It improves the visualization of small isolated
hot areas by enhancing the contrast (Fig.14.14). MIP is successfully employed
for lesion detection in PET oncological whole body studies. Its efficiency for the
detection of lesions is further increased by displaying the MIP projections as a
sequence of projection angles in cine mode.
Viewing plane (screen)

Splatting
Eye
Ray casting
Image data volume
FIG.14.13. Principle of ray casting and splatting geometry. In ray casting, the ray collects
intensity transformation throughout its trajectory. A voxel is usually hit outside its centre
which has to be corrected for by interpolation. Splatting starts from the centre of a voxel and
distributes its intensity on several screen pixels.
Summed voxel projection produces the rendered image by summing all

intensities along a ray trajectory. If applied to a CT volume, it is known as a
digitally rendered radiograph. If central projection geometry is used, the projection
image simulates a conventional X ray image. Digitally rendered radiographs of
CT data are used in radiotherapy for the positioning and registration of patients.
Tomographic datasets from nuclear medicine displayed as digitally rendered
535
CHAPtER 14
radiographs may be used to compare lesion extensions with planar X ray images
of the patient.
FIG. 14.14. A maximum intensity projection (right) compared to a standard coronal slice.
A solitary lesion is clearly visible in the maximum intensity projection image (arrow) while it is
missing in the coronal standard view.
14.7.2.2. Reflection type volume and surface rendering

The purpose of surface and volume rendering is to visualize structures in
volume datasets as objects and display them in a pseudo-3-d mode. Volume
rendering techniques extract information about objects directly from the 3-d
volume data. They start by casting rays through the image volume data and by
processing the intensities of the voxels along the ray trajectory. depending on the
handling of the intensities, different types of 3-d display can be generated.
Three dimensional rendering utilizes standard computer graphics
techniques, such as illumination, shading and the application of textures, to
produce a realistic appearance of anatomical structures and tumours. This is
useful for the visualization of complex anatomical relationships, can improve the
orientation of surgeons and resolve ambiguities of localization.
for registered images originating from different imaging techniques, such
as images from magnetic resonance and from PeT, anatomical and functional
536
data can be displayed simultaneously, thereby taking advantage of the excellent

morphological resolution of one modality and of functional, blood flow or
metabolic information of the other modality. Such combined images are
capable of displaying the spatial relationships between different objects, such
as, for example, the surface of the left ventricle together with the location of
the coronary arteries, or the surface of the brain grey matter rendered from a
magnetic resonance study combined with the blood flow obtained by an HMPAO
(hexamethylpropyleneamine oxime) SPECT study.
Surface rendering traditionally starts from a sequence of contours extracted
from the object of interest. The surface is obtained by fitting a mosaic of triangles
followed by illumination and shading. The relatively small number of parameters
needed to describe an object permits real time visualization of transformations,
useful, for example, for interactive surgical planning. Analytical descriptions of
the objects can also be generated, which can be used by other programs, such as
CAD/CAM, or which can be used to produce physical models of the objects of
interest using lithographic techniques.
Three dimensional surfaces are generated by specifying an intensity
threshold. The method is closely related to the generation of contours. When a
ray encounters the threshold value along its trajectory, the location of that voxel
is interpreted as a surface point of the structure of interest. The appearance of a
3D image is improved further by utilizing illumination and shading techniques.
In order to apply these effects, additional knowledge about the orientation of the
surface element is required for which gradient techniques with various levels of
sophistication are employed.
Voxel gradient shading is the most successful technique to produce
illuminated and shaded surfaces. It calculates a gradient vector from a voxel
neighbourhood and renders a realistic pseudo-3D image by calculating diffuse
reflective illumination from an external light source and applying shading. Noise
in the surface is reduced by smoothing (Fig.14.15, middle).
Volume compositing can be considered a generalization of voxel gradient
shading. It aims at visualizing internal structures beyond the limit given by a
threshold by using information from all voxels along a ray. The method consists
of calculating a gradient for each voxel along the ray, thereby assigning a
surface to each voxel, and applying lighting and shading to that surface. The
light transmitted and reflected by each voxel is then collected into the pixel
value on-screen by assigning opacities to each voxel and summing the results
along the ray (Fig.14.15, right). Volume compositing is by far the most complex
and time consuming rendering method. The results are similar to the voxel
gradient method. Under favourable conditions and through careful selection
of the rendering parameters, volume compositing can visualize several objects
simultaneously in the rendered image.
537
CHAPTER 14
FIG.14.15. Surface of a skull from CT image data using (from left to right) maximum intensity
projection, voxel gradient shading and volume compositing for rendering. Rendered images
were produced with ANALYZE 9.0.
14.7.3. Polar plots of myocardial perfusion imaging

Myocardial perfusion imaging is a tomographic technique using a
myocardial perfusion tracer such as the potassium analogue 201Tl or 99mTc-MIBI
(methoxyisobutylisonitrile) to produce SPECT images of the perfusion of the left
ventricle. Myocardial perfusion is reduced or absent in ischaemic and infarcted
areas. The size and severity of perfusion defects is of high diagnostic and
prognostic value. Therefore, myocardial perfusion imaging is among the most
frequent nuclear medicine investigations. Unfortunately, visual interpretation of
the tomographic slices is difficult and suffers from high inter-observer variability
due to the poor spatial resolution of SPECT studies in general and to the added
blurring of the images by the motion of the heart during acquisition. Display
methods tailored to a more reliable detection of perfusion defects were, therefore,
developed shortly after the introduction of myocardial perfusion SPECT.
The initial visual presentation of the tomographic images makes use of a
coordinate system natural to the anatomy of the left ventricle. One coordinate
axis passes through the long axis of the heart; the other two are perpendicular
to the long axis and to each other (Fig.14.16). The standard display consists
of slices perpendicular to the long axis, the short axis slices and two sets of
slices parallel to the long axis. The left ventricle has an annular shape in the
short axis slices. The annuli can be aggregated into one image using a polar map
representation [14.10]. In a first step, each annulus is reduced to a circumferential
profile. The methods of choosing the circumferential profile vary, for example,
using the maximum intensity at each angular step only or taking a mean intensity,
and the thickness of the annulus can also be considered at each angular step.
Thereafter, all of the profiles are arranged into one image, starting with the
538
profile representing the apex at the innermost position, with each following
profile surrounding the previous one. The resulting display is referred to as a
bulls-eye display or polar map (Fig.14.17). The latter name refers to the fact that
the intensity along a given annulus can easily be handled by a polar coordinate
system. The intensities displayed for each annulus correspond to the myocardial
perfusion in that slice or a segment thereof. Absolute perfusion values cannot be
derived from the intensities. The method to obtain an estimate of the degree of
hypo-perfusion and of the location of the perfusion defect consists of comparing
the relative intensity values in different segments of the annuli to the maximally
perfused segments of the same patient, and then comparing the pattern of relative
perfusion of the individual study with normal perfusion patterns. This permits an
estimate of both the degree and extent of the perfusion defects as well as a good
anatomical allocation to the coronary arteries causing the hypo-perfusion.
FIG.14.16. Results of a stressrest perfusion study with 99mTc-MIBI (methoxyisobutylisonitrile)

with the orientation of slices adapted to the long axis of the heart. Images show hypo-perfusion
of the inferior wall. The upper rows are stress images; the lower rows are images at rest.
(Courtesy of B. Knig, Hanuschkrankenhaus, Vienna.)
539
CHAPTER 14
FIG.14.17. Bulls-eye displays of myocardial SPECT perfusion studies. Normal perfusion (left)
and hypo-perfusion of inferior wall (right). The colours indicate the degree of perfusion, with
white normal, orange acceptable, red hypo-perfused, and green no perfusion. Also
indicated are the perfusion areas for the main coronary vessels LAD, LCX and RCA. (Courtesy
of B. Knig, Hanuschkrankenhaus, Vienna.)
14.8. DUAL MODALITY DISPLAY

Several techniques have been developed to display registered images
originating from different modalities, such as from a PET/CT study. The simplest
technique is to display the images belonging together side by side. Anatomical
information can be gained easily by using the linked cursor technique. The
CT image which has superior spatial resolution is, thus, used to determine the
anatomical location of a lesion visible in the PET image. The linked cursor
technique, while providing precise anatomical information, is impractical if
several lesions are present in the image as often is the case in oncological studies.
In such situations, alpha blending is helpful, which combines both the CT and the
PET image into one composite image.
Alpha blending consists of adding the images pixelwise with different
weight. The weight is called the transparency factor , with 0 1. The
composite pixel ICS is given by:
ICS(m, n)=IBG(m, n) + (1 )IFG(m, n) (14.4)
where
IBG is the intensity of the background pixel;
and IFG is the intensity of the foreground pixel.
540
When using the native grey scale images for both modalities, it is difficult
to distinguish clearly which intensity comes from which modality. The composite
display becomes much easier to interpret if one of the images uses a CLUT. In
this case, the formula has to be applied to each colour component separately:
RCS(m, n)=RBG(m, n) + (1 )IFG(m, n) (14.5)
GCS(m, n)=GBG(m, n) + (1 )IFG(m, n)
(14.6)
BCS(m, n)=BBG(m, n) + (1 )IFG(m, n) (14.7)

where
R, G and B are the colour components of the background image;
and I is the grey value of the foreground image.
In PET/CT alpha blending, the background image is usually the PET image
whereas the CT image as the foreground image retains the grey scale (Fig.14.18).
The composite display can be further improved by changing thresholds and
windows for each modality separately and interactively. For CT, the traditional
windows are usually employed.
FIG.14.18. PET/CT fused image display with a PET image on the left showing a hot lesion
at the border between the lung and rib cage. The fused image in the middle shows the location
inside the lung close to the pleura; the CT image on the right confirms and permits close
inspection of the position. Linked cursors point to the position of the lesion. The transparency
factor is 0.5.
14.9. DISPLAY MONITOR QUALITY ASSURANCE

Several standards define the performance parameters of display systems
(see, for example, Refs [14.11, 14.12]). The Report of task group 18 of the
American Association of Physicists in Medicine offers an exhaustive up to date
541
CHAPTER 14
set of performance parameters for current medical display monitors, together

with procedures and accompanying test images to assess performance [14.13]. It
contains limits of acceptance for all parameters, distinguishing between primary
devices and secondary devices.
In addition to quality assurance aimed at the individual display device, a
major component of quality assurance is to ensure a consistent display of the
image at all display workstations of a PACS, including different ambient light
conditions. This is resolved by including the calibration and validation of the
DICOM GSDF into the quality assurance framework.
Quality control at regular intervals of a display device is required because
the performance may change over time, due to ageing processes of the display
device, both for CRT and LCD displays, and due to changes in environmental
lighting with time.
14.9.1. Acceptance testing
The purpose of acceptance testing is to ensure that the performance
of purchased equipment complies with the specifications established in the
sales contract. The user should clearly specify in the contract the required
performance, the test procedures to assess the performance parameters and the
limits of acceptability. Reference [14.13] lists a set of performance parameters
which completely characterize the performance of a soft copy display device.
These are summarized in Table14.2.
For each of the parameters, several tests at various levels of sophistication
are described in detail. Most of the parameters can be assessed visually by
analysing the test images listed in Table14.2, possibly with additional use of
templates on transparency sheets, such as for the assessment of distortions. An
exhaustive set of test images has been made electronically available for these tests,
both in Joint Photographic Experts Group (JPEG) and DICOM format [14.3]. For
quantitative tests, such as for the calibration of luminance characteristic curves, of
the DICOM GSDF and of chromaticity values, luminance meters and colorimeters
with computer readout of the measured values and special software are necessary.
14.9.2. Routine quality control
In order to make sure that a display system meets the expected performance
during its economic lifetime, assessment of performance parameters at regular
intervals is necessary. Reference [14.13] recommends that a subset of the tests
in Table14.2, namely geometric distortions, reflection, luminance response,
luminance dependencies, resolution and chromaticity, be performed at
monthly to quarterly intervals, depending on the monitors stability. Tests for
542
TABLE14.2. PERFORMANCE PARAMETERS OF DISPLAY MONITORS

AND EQUIPMENT FOR MEASUREMENT
(according to Ref.[14.13])
Major required tools
Test
Equipment
Patterns
Luminance response
Luminance and illuminance meters
TG18-LN
TG18-CT
TG18-MP
Luminance dependencies
Luminance meter
TG18-UNL
TG18-LN
TG18-CT
Reflection
Luminance and illuminance meters
TG18-AD
Resolution
Luminance meter, magnifier
TG18-QC
TG18-CX
TG18-PX
Geometric distortions
Flexible ruler or transparent template
TG18-QC
Noise
None
TG18-AFC
Veiling glare
Baffled funnel, telescopic photometer
TG18-GV
TG18-GVN
TG18-GQs
Chromaticity
Colorimeter
TG18-UNL80
geometric distortions and for resolution are more important for CRTs, whereas
the dependence of luminance on the viewing angle is important only for LCD
displays.
In addition, Ref.[14.13] recommends that a daily check prior to clinical
work be performed by the user. It consists of evaluating anatomical test images
or a suitable geometrical test image such as a TG18-QC test image (Fig.14.19)
to verify adequate display performance. The instructions for assessing the quality
of a display device when using the TG18-QC test pattern are given in Box 14.1.
543
CHAPTER 14
FIG.14.19. Test pattern TG18-QC suitable for daily quality control of display monitor
performance using visual inspection [14.13].
Box 14.1. Instructions for visual assessment of image quality using the
TG18-QC test pattern as part of daily quality control by the user [14.23]
1. G
eneral image quality and artefacts: Evaluate the overall
appearance of the pattern. Note any non-uniformities or
artefacts, especially at black-to-white and white-to-black
transitions. Verify that the ramp bars appear continuous
without any contour lines.
2. G
eometric distortion: Verify that the borders and lines of the
pattern are visible and straight and that the pattern appears
to be centered in the active area of the display device. If
desired, measure any distortions (see section 4.1.3.2).
3.
Luminance, reflection, noise, and glare: Verify that all
16 luminance patches are distinctly visible. Measure their
luminance using a luminance meter, if desired, and evaluate the
results in comparison to GSDF (section 4.3.3.2). Verify that
the 5% and 95% patches are visible. Evaluate the appearance
of low-contrast letters and the targets at the corners of all
luminance patches with and without ambient lighting.
4. R
esolution: Evaluate the Cx patterns at the center and corners
of the pattern and grade them compared to the reference score
(see section 4.5.3.1). Also verify the visibility of the
line-pair patterns at the Nyquist frequency at the centre and
corners of the pattern, and if desired, measure the luminance
difference between the vertical and horizontal high-modulation
patterns (see section 4.5.3.1).
544
The most frequent influence on reporting comes from changes in ambient

lighting. An increase in the level of ambient light results in poorer discrimination
of structures in the darker parts of the image. This has to be compensated for by
adapting the GSDF to the current light level. Modern medical display monitors,
therefore, provide automatic measurement and recalibration features. A typical,
high performance LCD display used as a primary device includes a luminance
meter covering a small area of the monitor to continuously control the display
characteristic curve and the maximum brightness level. A second photometer
records the average ambient lighting. With such an arrangement, it is possible
to adjust the GSDF continuously to the DICOM required luminance values,
taking into account the changes in Lmin and Lmax, theminimum and maximum
luminance values including the luminance Lamb added by ambient light.
As an annual quality control of a display device, the TG18 working group
recommends performing all tests carried out during acceptance [14.13].
REFERENCES
[14.1] NATIONAL ELECTRICAL MANUFACTURERS ASSOCIATION, Digital Imaging
and Communications in Medicine (DICOM), Part 14: Grayscale Standard Display
Function, Rosslyn, VA (2003).
[14.2] INTERNATIONAL COLOR CONSORTIUM, Color Management, UK (2003),
http://www.color.org/slidepres2003.pdf
[14.3] NATIONAL ELECTRICAL MANUFACTURERS ASSOCIATION, Digital Imaging
and Communications in Medicine (DICOM), Supplement 100: Color Softcopy
Presentation State Storage SOP Classes, DICOM Standards Committee WG1DNEMA,
Rosslyn, VA (2005).
[14.4] PIZER, S.M., CHAN, F.H., Evaluation of the number of discernible levels produced
by a display, INSERM 88 (1979) 561580.
[14.5] SMITH, T., GUILD, J., The C.I.E. colorimetric standards and their use, Trans. Opt.
Soc. 33 (1931) 73134.
[14.6] INTERNATIONAL COLOR CONSORTIUM, The Role of ICC Profiles in a Colour
Reproduction System (2004).
[14.7] SONKA, M., HLAVAC, V., BOYLE, R., Image Processing, Analysis, and Machine
Vision, Brooks/Cole Publishing Company, Pacific Grove, CA (1999).
[14.8] ImageJ, A public domain Java image processing program, Version 1.32b (1997).
[14.9] BIRKFELLNER, W., et al., Wobbled splatting a fast perspective volume rendering
method for simulation of X-ray images from CT, Phys. Med. Biol. 50 (2005) N73N84.
[14.10] GARCIA, E.V., et al., Quantification of rotational thallium-201 myocardial
tomography, J. Nucl. Med. 26 (1985) 1726.
545
CHAPTER 14
[14.11] DINV 6868-57. DINV 6868-57: Sicherheit der Bildqualitt in rntgendiagnostischen

Betrieben, Teil 57: Abnahmeprfung an Bildwiedergabegerten, Normenausschu
Radiologie (NAR) im DIN Deutsches Institut fr Normung e.V. (2001).
[14.12] VIDEO ELECTRONIC STANDARDS ASSOCIATION, Flat Panel Display
Measurement Standard, Version 2, Milpitas, CA (2001).
[14.13] AMERICAN ASSOCIATION OF PHYSICISTS IN MEDICINE, Task Group 18
(TG18), Assessment of Display Performance for Medical Imaging Systems, AAPM
On-Line Report No. 03, College Park, MD (2005).
546
CHAPTER 15
DEVICES FOR EVALUATING IMAGING SYSTEMS
O. DEMIRKAYA, R. AL-MAZROU
Department of Biomedical Physics,
King Faisal Specialist Hospital and Research Centre,
Riyadh, Saudi Arabia
15.1. DEVELOPING A QUALITY MANAGEMENT SYSTEM
APPROACH TO INSTRUMENT QUALITY ASSURANCE
A quality management system (QMS) has three main components:
(a) Quality assurance (QA);
(b) Quality improvement;
(c) Quality control (QC).
The aim of a QMS is to ensure that the deliverables meet the requirements
set forth by the users. The deliverables can be, in general, all the services provided
in a nuclear medicine department, and the diagnostic imaging services in
particular. In this section, the primary focus is the diagnostic imaging equipment
and images produced by them.
15.1.1. Methods for routine quality assurance procedures
QA is a systematic programme for monitoring and evaluation of the process
of production. It is an all-encompassing management plan to ensure the reliability
of the production system. QA in diagnostic imaging, however, can help minimize
the uncertainties and errors in equipment performance by supervising the entire
image production process. This, in turn, will guarantee that the images generated
are of diagnostic quality. QA can also help identify and rectify the problems,
errors and malfunctioning and drifting of the performance earlier. Moreover, a
QA programme can help the standardization of the image production process
across centres and, thus, allows comparison of clinical results with other centres.
This is especially imperative in multicentre clinical trials. A QA programme in
nuclear medicine involves all aspects of nuclear medicine, including minimizing
the exposure to personnel, patients and the public; preparation, safety, sterility
547
CHAPTER 15
and administration of radiopharmaceuticals; patient handling; and ensuring the

diagnostic quality of images produced.
QC is the process by which the performance level of a product is measured
and then compared against the existing standards or tolerance values. QC
activities are a subset of QA activities. QA focuses on the processes while QC
focuses on the product.
QC with regard to imaging systems may entail:
A series of performance measurements to assess the quality of the imaging
system;
Keeping the record of measurements;
Monitoring the accuracy and precision of the results;
Taking corrective actions in case the performance measurements are outside
the tolerance levels or above the predetermined action levels.
The items above require:
Defining the performance parameters to be measured;
Preparing written procedures as to how and by whom the measurements
should be carried out;
Establishment of the frequency of performance tests and expected results in
the form of tolerance and action levels;
Training the persons who perform these measurements;
Designing record forms (preferably electronic) to keep the measurement
values;
Logging and reporting all of the problems and actions taken.
Tolerance levels define the range within which the results are acceptable
while action levels define the range beyond which a corrective action is required.
The upper level of the tolerance range may concur with the lower level of the
action range. If the performance of the system is just outside the tolerance range,
an immediate corrective action may not always be needed and the imaging
system can still be used for patient scanning, but a close monitoring of the system
performance is critical in the following tests. Record keeping is critical and
essential for trending the performance parameters to monitor the system and to
intervene, when necessary, in an effective and timely manner.
Phantoms are indispensable tools for QC measurements. They are utilized
to evaluate diagnostic imaging systems, as well as for other reasons in radiation
protection, radiobiology and radiotherapy. The phantoms can be hot (containing
a known amount of radioactivity) or cold (containing no radioactivity) for
primarily measuring radiation interaction. Phantoms used in nuclear medicine
548
are usually injected with a radioisotope simulating a particular organ or tissue

structure containing a particular radiopharmaceutical, while X ray computed
tomography (CT) QC phantoms are employed to measure CT values of water
and/or other materials by simulating different tissue types. IAEA Human Health
Series Nos 1 and 6 [15.1, 15.2] include an extensive discussion of the QA for
positron emission tomography (PET) and PET/CT systems, and single photon
emission computed tomography (SPECT) systems, respectively.
The International Commission on Radiation Units and Measurements
(ICRU) in ICRU Report 48 [15.3] defines the phantom as a material object
that includes one or more tissue substitutes and is used to simulate radiation
interaction in the body. Furthermore, any material that simulates a body tissue in
its interaction with ionizing radiation is termed a tissue substitute [15.4].
The ICRU distinguishes the physical phantoms from what are usually
called software phantoms by defining them as phantoms and computational
models, respectively. In this chapter, the convention of the ICRU is followed for
consistency in the terminology and to avoid any potential misunderstanding that
the other naming conventions may lead to.
According to Ref. [15.3], phantoms can be grouped under three categories
with respect to their primary usage: dosimetric, calibration and imaging
phantoms. The dosimetric phantoms are used to measure absorbed dose while
calibration phantoms are employed to calibrate a particular photon detection
system such as a PET scanner to convert the number of detected photons to
actual activity per tissue volume. An imaging phantom is used for assessing
image quality or characterizing imaging systems. The ICRU further defines
three subcategories under the above three functional categories. These are body,
standard and reference phantoms. Body phantoms are built in the shape of a body
and consist of multiple tissue substitutes or organs. These phantoms are more
often referred to as anthropomorphic phantoms as they simulate the human body.
The anthropomorphic torso phantom, which is discussed later in the chapter,
consisting of liver, heart, spine and lung inserts, is used in nuclear medicine for
testing image quality and is an example of this category.
In this chapter, physical phantoms or simply phantoms and computational
models that have applications in nuclear medicine are discussed. Throughout this
chapter, many commercial phantoms are mentioned and are pictured in figures
for ease of understanding. This does not, however, constitute an endorsement of
these commercial products.
549
CHAPTER 15
15.2. HARDWARE (PHYSICAL) PHANTOMS

The use of phantoms dates back to the beginning of the 20th century. In the
1920s, water tanks and wax blocks were often used for X ray experiments and, to
this day, these materials are still in use in certain applications. In the 1960s, more
reliable tissue substitutes and sophisticated phantoms began to appear.
Today, phantoms are used in performing numerous tasks within the field of
diagnostic imaging and radiation therapy. This includes testing the performance
of imaging equipment, measuring radiation dosage during therapy, teaching
interventional image guided procedures and servicing equipment in the field.
Hardware phantoms are the indispensable tools for medical physicists
to enquire about or characterize medical imaging systems. These phantoms
provide the means to determine, not only qualitatively but also quantitatively, the
performance characteristics of medical imaging systems.
As compared to computational models, physical phantoms may be
advantageous in that data are acquired with an actual scanner and contain the
effect of the parameters that impact on the entire photon detection process.
One major disadvantage of physical phantoms, however, is the difficulty of
simulating the change of the activity in an organ in time. Although phantoms that
simulate cardiac motion, for instance, are available commercially or are being
developed by researchers in various institutions, in general, phantoms simulating
physiological processes such as breathing are difficult to build and are not widely
available.
In this section, the hardware phantoms that are used to measure the
performance characteristics of gamma cameras and PET scanners are discussed.
Some of these phantoms are also known as test phantoms. Their physical
characteristics are reviewed along with a brief description of their purpose
of use. Some practical suggestions are also provided about the preparation of
the phantoms that require injection of radioactivity. Although the focus of this
section is primarily on discussing the phantoms themselves, the positioning and
data acquisition requirements are also addressed. The analysis of the acquired
phantom data is not the subject of this chapter. For the analysis of gamma
camera and SPECT performance test data, please see Ref. [15.5] in which the
test methods suggested by the National Electrical Manufacturers Association
(NEMA) are discussed. The authors have also developed a software application
and made it publicly available free of charge [15.5].
15.2.1. Gamma camera phantoms
The gamma camera is the most widely used diagnostic imaging system
available in nuclear medicine departments. Owing to their physical characteristics,
550
gamma cameras require very close attention and, therefore, more frequent and a
larger number of tests than any other diagnostic imaging modality in radiology.
One of the important QC tests that has to be carried out daily on every gamma
camera is the uniformity test. This test shows the current status of the gamma
camera and allows monitoring of any possible deterioration in the performance
of the camera. It can also signal whether there has been any malfunctioning in
the detector elements, such as the photomultiplier tubes or the crystal, since the
last QC test was conducted. These assessments can be performed qualitatively or
quantitatively by a computer program.
15.2.1.1. Point source holders
This phantom is used to hold point sources that are employed in intrinsic
uniformity, resolution and linearity measurements. It is made up of lead and
its main purpose is to shield the walls, ceiling and personnel, and collimate the
radiation to the detector. Figure 15.1 shows a picture of a source holder. Copper
plates (12 mm thick) should be placed in front of the source holder to act as
absorbers and stop the low energy photons. When placed on the floor, source
holder height can be adjusted such that the point source is directed to the centre
of the detector under investigation.
FIG.15.1. Point source holders in a slanted position so that they can point to the detectors
from the floor.
15.2.1.2. 57Co flood sheets

Gamma cameras should also be tested extrinsically (collimator in place)
using a 57Co sheet source. The cost of 57Co sheet sources is relatively high and
they should be replaced every 2 years. It should be noted that new sheet sources
551
CHAPTER 15
may contain 56Co and 58Co impurities. These radionuclides have a shorter
half-life (77.234 and 70.86 d, respectively) than that of 57Co (271.74 d) and emit
high energy rays (>500 keV). If the impurities result in non-uniformities, the
sources can also be left to decay for a while before being used. It is advisable to
place the sheet source at a distance of 510 cm from the collimator during the
scan. Figure 15.2 shows a commercial 57Co flood source.
FIG.15.2. Picture of a 57Co flood source.
15.2.1.3. Fillable flood phantoms

Although 57Co flood sources are more convenient and easy to use, their
higher cost may be a factor affecting accessibility. If one cannot have access to
57
Co flood sources, then a fillable water phantom source is a good alternative.
These phantoms are available commercially but they can also be manufactured
in a machine shop from Perspex. The commercial ones are available in different
dimensions for different detector sizes. It is necessary to be careful in filling these
phantoms to prevent bubble formation, contamination of the outside surface of the
phantom or the workplace, and/or bulging of the phantom in the centre. Bulging
of the phantom and air bubbles formed in the phantoms can affect the uniformity
of the captured image. Depending on the size and volume of the phantom, around
370 MBq (10 mCi) of 99mTc activity will be sufficient to give a count rate of
20 kcounts/s in the image. The acquisition of the image is performed in the same
way as the 57Co flood sources.
552
15.2.1.4. Slit phantom

Slit phantoms are used to measure the intrinsic resolution of a gamma
camera detector. The phantom is made of a 3 mm thick lead mask consisting of
1 mm wide parallel slits that are 30 mm apart. Slit phantoms, which are usually
manufactured by the gamma camera vendors, vary in size to fit perfectly to
particular detectors. They are made in pairs to measure the intrinsic resolution
in the X and Y directions (see Figure 15.3). These masks are placed in the
closest possible proximity to the crystal covering its entire area. Measurement
is performed using a 99mTc point source centred at a distance more than five
times the largest dimension of the useful field of view (UFOV) of the crystal.
The activity of the point source is adjusted, so that the count rate is less than 20
kcounts/s.
FIG.15.3. Top: picture of the slit phantom designed for a cardiac camera whose field of view
is smaller than that of a typical gamma camera. Bottom: acquired images of the slit phantoms
for a typical gamma camera to measure the resolution in the Y (left image) and X (right image)
directions. The white vertical and horizontal lines denote the image of 1 mm slits.
553
CHAPTER 15
15.2.1.5. Dual-line source phantom and scattering medium

This phantom, suggested by NEMA NU 1-2007 [15.6], is used to measure
the extrinsic resolution of the system with and without a scattering medium.
It consists of two parallel line sources 1 mm in internal diameter and with
a centre to centre distance of 5 cm. The line sources are built so that they are
positioned 10 cm above the collimator. Figure 15.4 shows a simple custom built,
dual-line source phantom. The capillary tube shown as dark lines in the figure is
commercially available but a butterfly IV line can also be utilized.
FIG.15.4. A custom built, dual-line source phantom. On the left is the phantom positioned
on the detector, and on the right the same line sources are immersed in a scattering medium
consisting of sheets of Perspex.
The line is filled with 99mTc activity solution with a concentration of about
550 MBq/mL (15 mCi/mL) to achieve an adequate count rate when used with the
scattering medium. When measuring X and Y resolutions, the lines are placed
parallel to the Y and X directions, respectively. In both cases, one of the lines
should be positioned in the centre of the field of view (FOV). The acquired image
should have at least 1000 counts in the peak channel of the line spread function.
To measure the extrinsic resolution with scatter, the dual-line source is
embedded into Perspex sheets, 10 cm of which are placed between the collimator
and the line sources and 5 cm placed above the lines as seen in Fig. 15.4. The
Perspex sheets under the sources create a scattering medium and the ones above
a backscattering medium. For a perfect contact between the sheet and the line
sources, it is recommended to make two grooves, through which the lines run, in
one of the sheets to insert the two lines.
15.2.1.6. Bar phantom
The second most frequent QC test in nuclear medicine is the resolution
test performed with bar phantoms. Bar phantoms can be used to measure,
554
semi-quantitatively (i.e. visually), the extrinsic and the intrinsic resolution of a

gamma camera. Images of bar phantoms can also be useful for the qualitative
evaluation of the gamma camera linearity which is normally measured by the slit
phantom.
Bar phantoms are made of lead strips embedded into plastic and typically
arranged in four quadrants. The lead strips are radio-opaque, while plastic strips
are radio-lucent. Each quadrant has strips of different thickness. The rectangular
bar phantom image shown in Fig. 15.5 (middle) has four quadrants with strip
sizes of 2.0, 2.5, 3.0 and 3.5 mm, while the image on the right has four quadrants
with strips of sizes 3.2, 4.6, 6.3 and 10 mm. In the images of the bar phantoms,
displayed in grey colour maps, white lines correspond to the plastic strips while
black lines correspond to lead strips. In a bar phantom, the strips are separated
with the same distance as the strip width.
FIG.15.5. Left: picture of a typical four-quadrant rectangular bar phantom. Middle: image
of the left bar phantom acquired by an ECAM gamma camera. Right: image of a bar phantom
acquired with an ADAC FORTE gamma camera. Both images were acquired at a matrix size
of 512 512 and with a total count of 10 Mcounts.
In routine QC tests, normally performed weekly or biweekly, bar phantoms

are used for the visual assessment of the extrinsic resolution (collimator mounted),
together with a flood source discussed in the previous section. Normally, a low
energy high resolution, parallel-hole collimator is used during this test. The bar
phantom is first placed directly on the collimator and the flood source is placed
on top of the bar phantom. Since the gamma camera resolution is dependent on
the distance from the detector, operators should make sure that the bar phantom
and the collimator are in direct contact with each other. A 10 Mcount image of
the bar phantom is normally acquired and evaluated visually to check the detector
resolution and linearity.
When used for determining the intrinsic resolution, the bar phantom is
again placed on the detector without the collimator in place, and a 99mTc point
source is placed at a distance five times the largest dimension of the crystal away
from the bar phantom. As a rule of thumb, the intrinsic resolution of a detector
555
CHAPTER 15
in terms of the full width at half maximum (FWHM) of the line spread function
can be approximately determined as FWHM 1.7Sb, where Sb is the size of the
smallest resolvable bars.
15.2.1.7. Dual-line phantom for whole body imaging
This phantom is used to test the whole body resolution of a gamma camera
system. It consists of two parallel line sources which are 1 mm in internal
diameter and 10 cm centre to centre. Figure 15.6 shows a custom built, dual-line
phantom. The line is usually filled with 99mTc activity with a concentration of
about 370 MBq/mL (10 mCi/mL) to achieve an adequate count rate. During the
testing, the line sources are placed at a distance of 10 cm from both collimators.
When measuring the perpendicular resolution, the lines should be placed parallel
to the bed direction with one of them being in the centre of the bed. When
measuring the parallel resolution, the lines should be positioned perpendicular to
the direction of the bed movement. The whole body resolution is calculated from
the FWHMs of the line profiles extracted from the image of the dual-line sources.
FIG.15.6. Dual-line phantom for whole body resolution tests.
15.2.1.8. Planar sensitivity phantom

In a planar sensitivity test, the accuracy of the response of the detector to a
radioactive source of known activity is measured for the particular collimator. It
is suggested to use a Petri dish containing around 3 mm of water homogeneously
mixed with a suitable activity (around 40 MBq) of 99mTc. The activity should
be drawn into a syringe and then measured accurately in the dose calibrator.
After injecting the activity in the Petri dish, the residual activity in the syringe
556
should be measured. The residual activity is subtracted from the initial activity
to determine the net activity injected into the dish. This dish should be placed
at a distance of 10 cm from the face of the collimator. It is recommended to
acquire two images. The average count, in units of counts per megabecquerel per
second or counts per minute per microcurie, is determined to measure the planar
sensitivity of the system.
15.2.1.9. Multiple window spatial registration phantom: lead-lined point source
holders
A multiple window spatial registration test measures the cameras ability to
position photons of different energies. In this section, the phantom is discussed,
as described in Ref. [15.6], together with its preparation and the measurement
procedures. The details of the test conditions and test phantoms can be found
in Ref. [15.6]. A schematic drawing of the lead phantom is given in Fig. 15.7.
As suggested by NEMA, nine of these lead-lined source holders are placed on
the surface of the detector. The relative position of each holder is shown in the
drawing. Plastic vials, as seen in Fig. 15.7, can be used to hold the actual activity
of 67Ga (~711 MBq (200300 mCi) in each). Other acquisition parameters and
camera settings are given in Table 15.1.
50 mm
25 mm
5 mm
Liquid Ga-67
source
0.8*UFOV
0.4*UFOV
90 mm
25 mm
5 mm diameter
hole
UFOV
FIG.15.7. Multiple window spatial registration phantom lead-lined point source holders. On
the right is the top view of the point sources or source holders placed on the detector crystal.
The locations of the point sources are determined by multiplying the dimensions of the useful
field of view (UFOV) by 0.4 and 0.8. On the left is the cross-sectional view of the source holder
together with the source vial.
557
CHAPTER 15
Images of nine (or four) point sources of 67Ga are acquired normally at
three different photopeak energy windows (the three photopeaks for 67Ga are 93,
185 and 296 keV).
The aim of the subsequent calculation is to find the centroids of these
points in the image acquired at different energy windows and to compare the
displacement between the point source images acquired at different energy
windows. The maximum displacement between the centroids of point sources
is the performance parameter indicating the error in multiple window spatial
registration. The details of the calculation of this performance parameter can be
found in Ref. [15.6].
TABLE15.1. IMAGE ACQUISITION AND CAMERA SETTINGS
67
Radionuclide
Ga
activity
~710 MBq in each source
Total counts
1000 counts in the peak pixel of each point source
Energy window
15%
Count rate
<10 kcounts/s
Pixel size
<2.5 mm
Matrix size
~1024 1024
15.2.2. SPECT phantoms

15.2.2.1. Triple-point source for SPECT resolution
Triple-point source phantoms are used for measuring the SPECT resolution
in air (i.e. under no scatter conditions) or measuring centre of rotation (COR)
alignment. Further details on the test conditions and the phantom can be found in
Ref. [15.6].
For this purpose, thin-walled glass capillary tubes with an internal diameter
of less than 2 mm are used. These point sources can be prepared as follows.
First, a 99mTc solution of high concentration (about 5.5 GBq/mL) is prepared in
a small (1 mL) syringe. Then, drops of small sizes are created on the surface of
a clean plastic. These small drops can be drawn up into the capillary tubes by
the principle known as capillary action by simply touching them. It may take
a few trials to get a small size drop. At the end, the capillary tubes should be
sealed on both ends with a capillary tube sealer such as Critoseal. The point
sources should be made as spherical as possible, that is, their transaxial and
558
axial extents should be similar in length. Their maximum dimension (the axial
extent of the activity) should not exceed 2 mm. The activity in the point sources
should not vary more than 10%. The point sources should be suspended in air
and positioned in accordance with the suggestions in Ref. [15.6] (Fig. 15.8). An
alternative practical solution to suspend the point sources in air is to mark the
positions of the point sources on a thin paper attached to a polystyrene (widely
known as Styrofoam) sheet, and use this as a source holder. The scatter caused by
the holder should be negligible.
FIG.15.8. Top and side views of the position of the point sources as suggested by the National
Electrical Manufacturers Association.
15.2.2.2. Triple-line source phantom for SPECT resolution

The SPECT resolution with scatter is measured using the triple-line source
phantom. This performance test is normally performed as part of the acceptance
testing and annual testing. As described in Ref. [15.6], this phantom consists of a
cylinder made of plastic (lucite or Perspex) with three line sources oriented along
the axial direction (see Fig. 15.9). The cylinder is filled with water to create a
scattering medium. The line sources are available either as inserts of 57Co lines or
hollow metal tubes to be filled with 99mTc solution. Here, the latter is discussed
(Fig. 15.10). The inner diameter of the line sources is less than 2 mm. Both ends
of the line sources are available for injecting the activity and are normally closed
with small caps after the injection.
559
CHAPTER 15
Front view
Side view
75 5 mm
FIG.15.9. Schematic drawing of the front and side views of a triple-line source phantom.
FIG.15.10. A commercial triple-line source phantom with three line sources inside. The tank
is filled with water to simulate a scattering medium.
The line sources should all be emptied of the decayed solution left from the
previous test using two empty syringes attached at both ends of the line source.
During the injection of each line source, two syringes are attached to both ends,
one empty and one with activity of a concentration around 300500 MBq/mL.
While pushing the plunger of the syringe with the activity, that of the empty
syringe should also be pulled very slowly until 99mTc solution appears from the
other end. The filled line source should be securely sealed from both ends with
the original caps, ensuring that there is no leak. It should also be ensured that the
entire line source is uniformly filled.
560
During measurement, according to Ref. [15.6], the centre line source

should be on the axis of rotation centred in the FOV within 5 mm. The pixel size
should be small enough (<FWHM/3) to prevent aliasing. Since this test has to be
carried out with the collimator, a high resolution collimator is the best choice.
During data acquisition, a 0360 range should be evenly covered. Some of the
acquisition parameters and camera settings are given in Table 15.2.
TABLE15.2. ACQUISITION PARAMETERS AND CAMERA SETTINGS
FOR THE SPECT RESOLUTION WITH SCATTER TEST
Radionuclide
99m
Tc
Count rate (kcounts/s)
<20
Total kilocounts per view
100
Scan time/view
~5 s at 20 kcounts/s
Energy window
15%
Collimator
Radius of rotation
Total number of views
Pixel size
Low energy high resolution

150 5 mm
120
<2.5 mm
After the measurement, the resolution parameters should be calculated

according to the method set forth in Ref. [15.6].
15.2.2.3. Volume sensitivity and detector to detector variation measurement
phantom
Volume sensitivity is the total system sensitivity to a uniform concentration
of activity in a specific cylindrical phantom. Factors such as detector
configuration, collimator type, radionuclide, energy window setting and source
configuration will impact the volume sensitivity in SPECT. Detector to detector
sensitivity variation is the relative difference in sensitivity of the individual
detector heads in a tomographic mode. The data acquired in a volume sensitivity
test are directly used to calculate this performance parameter as well.
The volume sensitivity in SPECT is measured using a cylindrical phantom
with an inner diameter and a length of 200 5 mm (see Ref. [15.6]). The
recommended wall thickness is 10 2 mm. The volume of the phantom has to
be accurately measured to accurately calculate the source concentration. The
561
CHAPTER 15
phantom is filled with water uniformly mixed with a known amount of activity
(approximately 350 MBq) of 99mTc. The activity amount should be such that
the count rate at the photopeak energy window is 10000 2000 counts/s. The
following parameters have to be accurately determined and recorded to calculate
the volume sensitivity:
Volume of the phantom;
Pre- and post-injection syringe activity to determine net injected activity;
Elapsed time half way through the SPECT acquisition;
Total scan time.
Further details of the measurement and calculations can be found in
Ref. [15.6].
15.2.2.4. Total performance test phantoms
Image quality measures or overall SPECT system performance, such as
noise, tomographic uniformity, contrast and lesion detectability, are measured
using total performance phantoms. These phantoms are commercially available
and are not so easy to build in an institutional workshop. There are several
commercial phantoms for this purpose. Some of the phantoms that are frequently
used to assess the performance of a SPECT system are discussed. It should be
noted that these phantoms can also be used to evaluate PET systems.
15.2.2.5. Carlson phantom
The Carlson phantom (designed and developed by R.A. Carlson, Hutzel
Hospital, Detroit, MI, USA, and J.T. Colvin, Texas Oncology PA, Dallas,
TX, USA) in this category is frequently used for evaluating the tomographic
uniformity, image contrast, noise and linearity. The main source tank
(see Fig. 15.11) is made of acrylic with dimensions: 20.32 cm inside diameter,
21.59 cm outside diameter and 30.48 cm length. The phantom comes with
various inserts, which are demonstrated and described in Fig. 15.11, to evaluate
the performance parameters noted above. The thick plastic screws on the top lid
allow easy filling and draining of the tank with water. The 99mTc solution injected
inside the tank serves as the background activity, which may vary between 300
and 550 MBq, depending on the collimator used [15.7].
There is an insert or section for each performance measure. The
SPECT uniformity is assessed using the uniform section of the phantom. The
non-uniformities in the gamma camera can result in severe ring or bulls-eye
artefacts. These artefacts can be checked for by looking at the uniform transverse
562
slices. The amount of noise can be quantitatively calculated from the uniform
section.
15.2.2.6. Jaszczak circular and elliptical phantoms
Similar to the Carlson phantom, Jaszczak elliptical and circular phantoms
are used to evaluate the overall performance of SPECT systems after a repair
or preventive maintenance, or during acceptance testing or quarterly testing. In
addition to the purposes above, these phantoms can be used in evaluating the
impact of reconstruction filters on resolution, as well as for other purposes in
research studies.
Jaszczak phantoms consist of a main cylinder or tank made of acrylic
with several inserts (see Fig. 15.12). They are manufactured and sold by Data
Spectrum Corporation (NC, USA). Jaszczak phantoms, which may have circular
or elliptical tanks, come in several different flavours. The cylinders of all
models of the circular flanged phantoms have the same physical specifications:
21.6 cm inside diameter, 18.6 cm inside height and 3.2 cm wall thickness. The
principal differences between the different models of the flanged cylindrical
Jaszczak phantoms are the diameters of the rods and solid sphere inserts. The
circular phantom has flanged and flangeless models. The latter is recommended
by the American College of Radiology for accreditation of nuclear medicine
departments. These different models are designed to test a range of systems, from
low resolution to ultra-high resolution, which has rods and spheres smaller than
the others.
All Jaszczak phantoms have six solid spheres and six sets of cold rods. In
flanged models, the sizes of the spheres vary. The number of rods in each set
depends on the size of the rod in that set as different models of the phantom have
rods of different sizes. In flangeless models, the diameters of the spheres are 9.5,
12.7, 15.9, 19.1, 25.4 and 31.8 mm, while the rod diameters are 4.8, 6.4, 7.9, 9.5,
11.1 and 12.7 mm. Both solid spheres and rod inserts mimic cold lesions in a hot
background. Spheres are used to measure the image contrast while the rods are
used to investigate the image resolution in SPECT systems.
15.2.2.7. Anthropomorphic torso phantoms
Anthropomorphic torso phantoms are used in testing gamma cameras in
SPECT mode to evaluate data acquisition, attenuation correction and image
reconstruction methods. They normally simulate or model the upper torso of
the body (from the heart down to the diaphragm) of an average male or female
patient. These phantoms consist of a body-shaped (elliptical) cylinder with
fillable inserts for organs such as the heart, lungs and liver (see Fig. 15.13).
563
CHAPTER 15
Phantom
Description
Hot lesions
Eight pairs of holes drilled through a solid

acrylic block, with diameters of 4.7, 5.9, 7.3,
9.2, 11.4, 14.3, 17.9 and 22.3 mm, model hot
lesions with the background activity injected.
Cold rods and spheres

Seven rods, with diameters of 5.9, 7.3, 9.2,
11.4, 14.3, 17.9 and 22.3 mm, simulate cold
lesions. Each rod is 25% larger in diameter
than the preceding one. Seven solid spheres
of the same diameters as rods, the centre one
being the largest, are attached to the rods.
Linearity/uniformity section
Crossed grid of cut out channels, again in
an acrylic block, can be used to assess the
linearity. The region where only background
activity is available is used to evaluate the
tomographic or SPECT uniformity.
Picture of the Carlson phantom tank together

with all three inserts.
FIG.15.11. Carlson phantom and its inserts.
564
FIG.15.12. Jaszczak phantom used for verifying image quality (phantom by Data Spectrum
Corporation, USA).
Defects can also be added to the heart insert. Lung inserts are filled with
Styrofoam beads and water to emulate lung tissue density. The phantoms can be
used to evaluate non-uniform attenuation correction methods including CT based
attenuation correction in SPECT/CT systems and scatter compensation methods.
When used with the optional cardiac insert, cardiac SPECT data acquisition and
reconstruction methods may also be evaluated.
Filling the inserts with different distributions of radioactivity is not as
easy as filling other phantoms because of the multiple organs and the organ
to background ratios that need to be adjusted. To set the concentration ratios,
the volumes of the organ inserts need to be measured accurately a priori. For a
simulation of a 1110 MBq (30 mCi) sestamibi stress study, the injected activity
concentrations, as suggested in Ref. [15.8], are given in Table 15.3.
Torso phantoms can be integrated with the fillable breast phantom, which
is also commercially attainable. These breast phantoms allow the inclusion
of inserts to simulate breast lesions that can be employed to evaluate lesion
detectability.
The volumes in the second column in Table 15.3 are the measured volumes
of the torso phantom inserts.
565
CHAPTER 15
TABLE 15.3. SUGGESTED ACTIVITY CONCENTRATIONS AND

MEASURED VOLUMES OF INSERTS FOR THE ANTHROPOMORPHIC
TORSO PHANTOM
Volume (mL)
Activity concentration
(kBq/mL)
Total activity
(MBq)
Heart
117
250
30
Tissue
8620
25
225
Liver
1177
150
175
Section
Lungs
FIG. 15.13. A commercial anthropomorphic phantom and a transaxial slice cutting through
the heart and lungs from its image acquired by a SPECT/CT system.
15.2.2.8. Hoffman brain phantom

This phantom, developed by Hoffman et al. [15.9], provides an anatomically
accurate simulation of the radioactivity distribution in normal brain. Using this
phantom, cerebral blood flow and metabolic activity in the brain can be simulated.
It can be used in both PET and SPECT systems to optimize/investigate imaging
acquisition protocols, to evaluate attenuation and scatter correction methods, and
to measure the performance of imaging systems. It consists of a water fillable
cylinder (i.e. a single-fillable chamber) containing 19 separate layers each 6.4 mm
thick (see Fig. 15.14). The fillable water volume is about 1.2 L. Water freely
permeates between layers to simulate concentration ratios of 4:1:0 between grey,
white and ventricle, respectively, in normal brain. The 2-D version, consisting of
a single slice, and a 3-D version of the phantom are available commercially.
566
DEVICEs FOR EVALUAtInG IMAGInG sYstEMs
FIG. 15.14. Three dimensional Hoffman phantom with a water fillable cylinder and layers of
inserts (phantom by Data Spectrum Corporation, USA).
15.2.2.9. Defrise phantoms

These phantoms are designed for measuring the performance of small animal
imaging systems (both sPecT and PeT). They can be used to investigate image
quality or resolution. figure 15.15 shows the hot spot phantom manufactured by
data spectrum corporation, usa. This phantom is a miniaturized version of the
image quality phantoms mentioned in the previous sections. The phantoms are
available in different sizes for imaging systems with different foVs.
FIG. 15.15. Defrise hot spot phantom manufactured by Data Spectrum Corporation, USA.
567
CHAPTER 15
15.2.3. PET phantoms

15.2.3.1. National Electrical Manufacturers Association image quality phantom
Measuring image quality in an objective manner has been one of the most
difficult tasks in PET. Image quality in PET can be determined by calculating
performance parameters, such as uniformity, noise, lesion contrast, spatial
resolution, and the accuracy of the attenuation and scatter correction techniques.
In this section, the NEMA image quality (IQ) phantom is described. This phantom
(known as the NEMA IEC (International Electrotechnical Commission) body
phantom) was originally recommended in IEC standards and was then adopted by
NEMA. In addition to the above performance parameters, the image registration
accuracy between the PET and CT gantries in a PET/CT scanner can be assessed.
This phantom is commercially available from Data Spectrum Corporation, USA.
The IQ phantom consists of four main parts:
(a) Fillable spheres: The six fillable spheres are used for measuring hot and
cold lesion contrast. The inner diameters of the six spheres are 10, 13, 17,
22, 28 and 37 mm. The two largest spheres (28 and 37 mm) are filled with
water to mimic cold lesions, while the rest are injected with 18F activity
with lesion to background ratios of 4:1 and 8:1 to mimic hot lesions. The
spheres are attached to the cover or top lid through capillary stems. The
FIG.15.16. Cross-section of the body part of the International Electrotechnical Commission

image quality phantom made of acrylic. The dimensions are given in millimetres (reproduced
with permission).
568
FIG.15.17. Transaxial (top left) and coronal (bottom left) cross-sectional view of the image
quality (IQ) phantom through the centres of fillable spheres. Sphere diameters and the other
dimensions are given in millimetres (reproduced with permission). On the right is the schematic
drawing demonstrating the positioning of the IQ phantom together with the scatter phantom.
filling is also done through the capillaries without removing the cover lid.
Filler screws for each fillable part inside the body phantom allow easy
access. A picture of the phantom is shown in Fig. 15.18.
(b) Cylindrical insert: A cylindrical section that is filled with a mixture of
polystyrene beads and water to mimic lung (the density of which is around
0.3 0.1 g/mL) attenuation is placed axially in the centre of the phantom
with the same length as the body phantom. The outside diameter of the
insert is about 5 cm.
(c) Phantom preparation: Table 15.4 shows the measured volumes of the
various inserts and the torso cavity of the IQ phantom. It is suggested that all
the volumes be measured upon acquiring a new IQ phantom. The activities
used to fill the phantom should be measured using a calibration time that
corresponds to the planned PET acquisition time, taking into account the
time necessary for the preparation and positioning of the phantom for
this test. Table 15.4 shows the typical activity concentrations that may be
prepared and injected into the background and the hot spheres in order to
have the proper activity concentration at the time of the scan (supposed to
be performed 45 min after phantom preparation). It should be noted that the
activity concentration ratio in the table is 8:1. A 4:1 activity concentration
ratio can be easily obtained by doubling the amount of activity in the
background.
569
CHAPTER 15
FIG.15.18. National Electrical Manufacturers Association/International Electrotechnical

Commission image quality phantom.
TABLE 15.4. MEASURED VOLUMES OF THE NATIONAL

ELECTRICAL MANUFACTURERS ASSOCIATION/INTERNATIONAL
ELECTROTECHNICAL COMMISSION IMAGE QUALITY PHANTOM AND
SUGGESTED ACTIVITIES FOR A CONCENTRATION RATIO OF 8:1
Volume (mL)
Typical
activity
(MBq)
9700
n.a.
Four hot spheres
Different sizes
n.a.
Two cold spheres
Different sizes
n.a.
Lung insert
353
n.a.
Background
(torso all inserts)
9286
65
Phantom section
Torso cavity
Activity
concentration
at time of
preparation
(kBq/mL)
Activity
concentration
at time of
scan
(kBq/mL)
56
42.4
5.3
Note: n.a.: not applicable. The scan is supposed to be performed 45 min after phantom
preparation. For a description of the phantom, please see: http://www.spect.com/pub/
NEMA_IEC_Body_Phantom_Set.pdf
570
There are different suggestions as to how to prepare the IQ phantom. The

following is a summary of one possible approach:
The NEMA recommends an activity concentration for the background of
5.3 kBq/mL at the time of the scan, assuming that a normal 70 kg patient
injected with 370 MBq of activity will have a similar background activity
in the body (370 MBq/70 000 mL, ~5.3 kBq/mL).
The amount of time to fill and position the phantom must be estimated to
determine the amount of activity at the time of preparation of the phantom.
A typical time frame for this process would be 45 min.
Two separate activities of 65 MBq are prepared and one of them is injected
into the background. This results in a background activity concentration of
65 MBq/9286 mL = 7 kBq/mL. The activity concentration will be reduced
to ~5.3 kBq/mL after 45 min (time of scanning).
Another solution for the hot spheres with an activity concentration of ~56
kBq/mL is prepared separately. 18F activity of 5.6 MBq can be injected into
100 mL of cold (non-radioactive) water to obtain this concentration. If the
measured activity is slightly more or slightly less, the volume of the cold
water can be adjusted accordingly to achieve the intended concentration.
The 18F activity is injected into the torso cavity (i.e. background), which is
already filled with cold water, and then the hot spheres are filled with the
prepared 18F solution.
After having acquired the images of the phantom for the ratio of 8:1, the
previously prepared activity of 65 MBq is added to the background in order
to obtain an activity concentration ratio of 4:1.
The phantom is acquired for the 4:1 ratio one half-life (~110 min) after the
first scan, when the activity concentration in the background will be ~5.3
kBq/mL.
One of the disadvantages of the above filling method is the difficulty of
mixing the background activity uniformly into the cold water; however, the
method obviates the need for removal of the top lid with attached spheres and
refilling of the background in each experiment.
15.2.3.2. National Electrical Manufacturers Association scatter phantom
The scatter phantom, whose specifications were set forth by NEMA
guidelines (NEMA NU 2-2007 [15.10]), is used to measure the count rate
performance of PET scanners in the presence of scatter. In other words, it is used
to measure the amount of scatter in terms of the scatter fraction, the effect of dead
time and the random events generated at different levels of source activity.
571
CHAPTER 15
The phantom consists of a solid, 70 cm long, polyethylene cylinder with

an outer diameter of 203 3 mm and a line source insert. The line source insert
is made of a clear polyethylene tube at least 80 cm in length, and with inner and
outer diameters of 3.2 0.2 and 4.8 0.2 mm, respectively. The volume of the
line source is approximately 6 mL. The solid cylinder comes in four segments for
ease of fabrication and handling. During the assembly, these four segments should
be tightly fitted to prevent the formation of scatter-free air gaps in between them.
A hole (6.4 0.2 mm in diameter) is drilled along the central axis of the cylinder
at a radial distance of 45 1 mm (see Fig. 15.19) to insert the aforementioned
line source. The scatter phantoms are commercially available.
The line source insert should be uniformly filled with a 18F solution. The
amount of activity is usually recommended by the manufacturer and should be
in the central 700 5 mm part of the insert. The line source should be inserted
such that the activity region remains completely within the 70 cm long phantom.
Further detail about phantom preparation and data acquisition can be found in
Ref. [15.10].
Centre of the phantom is located
at the centre of the field of view
Line source hole
FIG.15.19. Positioning of the scatter phantom on the patient bed: transaxial view (left);
picture of the National Electrical Manufacturers Association scatter phantom (right).
15.2.3.3. National Electrical Manufacturers Association sensitivity phantom

Sensitivity is the number of counts per unit time per unit of radioactivity
concentration within the FOV. To be able to compare different PET scanners, the
sensitivity performance measure should be independent of factors such as scatter,
attenuation, count losses and random events. Therefore, in PET, unlike SPECT,
the sensitivity is measured using a special phantom developed by Bailey et al.
[15.11] and later adapted by NEMA. The NEMA sensitivity phantom allows the
determination of the attenuation-free sensitivity.
572
The sensitivity phantom consists of five concentric aluminium sleeves

(70 mm in length), each with a wall thickness of 1.25 mm. The inner diameters
of the five tubes are 3.9, 7, 10.2, 13.4 and 16.6 mm. The line source, made from
clear polyethylene, is filled uniformly with 18F in solution and inserted into the
smallest sleeve and suspended in air within the FOV of the scanner. The line
source is filled with activity, such that the dead time losses are less than 1%
and the random events are less than 5% of the true rate. Figure 15.20 shows the
sensitivity phantom: the five aluminium sleeves and the tube. The figure also
shows the positioning of the phantom during the scan. In this case, a shower
curtain rod and the point source holder from the scanner vendor are used to
suspend the phantom within the FOV. A sling can be constructed from tape to
hang the phantom in that position as well. It should be noted that the centre of the
aluminium sleeves should coincide with the centre of the AFOV of the scanner.
FIG.15.20. Pictures of the National Electrical Manufacturers Association sensitivity

phantom: positioning of the phantom within the gantry (right). A spring tensioned shower
curtain rod and the point source holder are used to suspend the phantom within the field of
view. The aluminium sleeves (left and centre) should coincide with the centre of the axial field
of view.
15.2.3.4. Triple-point source phantom for spatial resolution

Hematocrit or capillary tubes are commonly used to create point sources
for measuring the spatial resolution of PET scanners. The inner and outer
diameters of these tubes should be less than 1 and 2 mm, respectively. The axial
extent of the activity in the tube should be no more than 1 mm. As for the NEMA
NU 2-2007 guidelines [15.10], three point sources should be positioned as shown
in Fig. 15.21. It should be noted that the central point source is positioned 1 cm
above the centre of the FOV.
A high concentration of 18F activity in a solution should be prepared such
that neither the dead time losses nor random events exceed 5% of the total event
rate. The actual activity concentration to be used, more than approximately
200 MBq/mL, is normally provided by the manufacturer. The preparation of the
573
CHAPTER 15
point sources in hematocrit tubes is undertaken as discussed in Section 15.2.2.1.

The point sources are positioned and the data are acquired at the centre of
the FOV as well as at a distance a quarter of the FOV away from the centre
(see Fig. 15.21). Figure 15.22 shows a point source holder with capillary tubes
mounted on it.
Centre of AFOV
10 cm
1 cm
Axial direction
10 cm
AFOV/4
Side view: AFOV

FIG.15.21. Positioning of the three point sources in the centre of the axial field of view
(AFOV). The view into the gantry bore (left) and the side view (right) in which the dashed
circles denote the axial position of the sources in the second scan.
FIG.15.22. Three capillary point sources mounted on a point source holder used in PET to
measure spatial resolution.
574
15.3. COMPUTATIONAL MODELS

Computational models can be categorized in three groups:
(a) Mathematical models;
(b) Voxelized computational models;
(c) Hybrid computational models.
This enumeration also reflects the progress of the development as the listing
is from simple to more realistic and sophisticated. This order of classification also
reflects the chronological order of the development process of the computational
models.
Mathematical models, also known as stylized models, simulate the organs
with geometric primitives such as ellipsoids, cylinders, spheres and rectangular
ellipsoids. These rather simple, geometrically well defined shapes representing
the organs or structures in the body are defined using the surface equations of
these primitives. The mathematical models were very early models and crude
in their representation of organs. The well known models, which have been
adopted by the Medical Internal Radiation Dose Committee of the Society of
Nuclear Medicine and have been used for many years in dose calculations, are
mathematical models. Although for a while these models served the purpose,
the need for a more realistic definition of organs, and, therefore, a more realistic
representation of the body, has always been there.
The advent of tomographic imaging technology, particularly X ray CT and
magnetic resonance imaging (MRI) made it possible to obtain high resolution
images of the body. In voxelized models, also known as tomographic models, the
organs are defined by the structures segmented from high resolution tomographic
images such as X ray CT and MRI. The segmented structures consist of
volumetric image elements called voxels, each of which is assigned a value
indicating the organ to which it belongs. The smaller the voxel dimensions, the
more realistic the surface of organs can look. Depending on the dimension of the
voxels, it may be challenging to define thin or small structures such as skin.
The Visible Human Project initiated and conducted by the United States
National Library of Medicine has played a significant role in the development of
voxelized models. As part of this project, CT and MRI images and cryosection
photographs of a 38 year old male cadaver were made available in the public
domain. To produce the colour photographic images of the cryosections, the
cadaver was frozen and sliced into 1 mm thin sections and photographed at a
resolution of 2048 pixels 1216 pixels. This project has led to the development
of many voxel based computational models [15.1215.14]. The construction of
voxel based models is a lengthy and tedious process and requires several steps.
575
CHAPTER 15
First, high resolution images of the body need to be acquired. Then, the individual
organs and structures are segmented from the high resolution images. The
segmentation is the most challenging task as the boundaries between organs and
tissues are often not well defined. Researchers, therefore, resort to tedious manual
or semi-automated segmentation methods. Obtaining CT scans of desired pixel
resolution or dimension and slice thickness may result in a significant amount
of exposure to ionizing radiation; thus, it is difficult to recruit healthy subjects
for this purpose. As a result, some of the voxel models have been constructed
from medical images of patients. For example, the Zubal phantom [15.15] was
created from CT scans of a patient by manual segmentation. These limitations on
pixel dimensions and slice thickness have made cadavers an attractive choice for
building voxel based models. In voxelized models, the surface of the organs are
jagged, piece-wise continuous and, therefore, not smooth. Other issues, such as
shifting of internal organs and non-rigid transformations in organ shape during
the scan in the supine position, may limit the generality of these models.
Hybrid models combine the best of both worlds. Surfaces of the segmented
structures in voxelized models are defined by mathematical formulations used to
define irregularly shaped surfaces such as 3-D B-spline surfaces.
A group of researchers developed a series of 3-D and 4-D computational
models. Their first model, the mathematical cardiac torso phantom, was a
mathematical model based on simple geometric primitives but also used
cut-planes and overlaps to create complex biological shapes to be used in nuclear
medicine research. This model also included a beating heart based on gated MRI
patient data and a respiratory model based on known respiratory mechanics. With
this model, emission and transmission data could be simulated. The following
models, 4-D NCAT and cardiac torso (XCAT) (see Fig. 15.23), were based on
the visible human CT dataset. The organ shapes, i.e. surfaces, were reconstructed
using the primitive non-uniform rational B-spline (NURBS) surfaces. The 4-D
models use cardiac and respiratory motions developed using 4-D tagged MRI
data and 4-D high resolution respiratory-gated CT data, respectively. These
models, from the hybrid class, can successfully model not only the anatomy but
also physiological functions such as respiratory and cardiac motion.
Such 4-D models can be used to accurately simulate SPECT and PET
images of the torso and can be particularly helpful for optimizing image
acquisition protocols and image reconstruction algorithms, and understanding the
various effects of these complex motions on the acquired PET or SPECT images.
These models are also accessible free of charge for academic research.
These models have been widely used in internal absorbed dose calculations
in nuclear medicine or in calculation of dose distribution from external sources
in radiation therapy and in studying issues pertinent to imaging systems and their
576
performance characteristics. They have also been quite helpful in the optimization
of image acquisition protocols and reconstruction methods.
FIG.15.23. Left: initial extension of the 4-D XCAT anatomy. Right: simulated chest X ray
CT images from the extended 4-D XCAT. Coronal (top row) and transaxial (bottom two rows)
reconstructed slices are shown (reproduced with permission from P. Segars).
Since anatomy and physiological functions are accurately known, they can
serve as gold standards. Computational models may be preferred because the use
of physical phantoms leads to unnecessary occupational exposure to radiation,
and the preparation and repetition of the experiments using physical phantoms
can be lengthy and time consuming.
An ideal model should be able to conform, reasonably well, to the size
and shape of the object being represented. Currently, as personalized medicine
is the strong driving impetus for most current research in many pertinent fields,
personalized modelling should be the aim in computational model development
research.
15.3.1. Emission tomography simulation toolkits
15.3.1.1. SimSET
SimSET, first released in 1993 and developed at the University of
Washington, is a simulation package that can simulate PET and SPECT emission
tomography systems using Monte Carlo simulations. It can model the photon
interaction process as well as the imaging detector geometries. SimSET allows
577
CHAPTER 15
the use of a different object description such as a Zubal phantom to simulate a

whole body phantom. SimSET is freely available for use.
15.3.1.2. GATE
Owing to the limitations of SimSET regarding the modelling of complex
detector geometries, the need for a more sophisticated emission tomography
simulator arose. To meet this need, a group of physicists from different institutions
around the world formed the OpenGate collaboration. Out of this collaboration, a
simulation toolkit (GATE) for nuclear medicine applications was developed and
has been available since 2001. GATE uses the existing libraries of Geant4, which
is a comprehensive simulation toolkit that simulates the interaction of particles as
they traverse through matter. GATE is unique and superior in that it can model
time dependent phenomena such as source and detector movement and source
decay kinetics. It includes validated geometry modelling tools that can model
complex scanner geometries. It also includes the description and models of
several commercially available PET and SPECT scanners. GATE can simulate
CT scans and can perform dose calculations. GATE is also freely available for
use.
15.4. ACCEPTANCE TESTING
As discussed in Ref. [15.16], gamma cameras are evaluated at different
levels of testing before being made ready for clinical use. The first set of tests is
carried out in the factory before shipment. Manufacturers test gamma cameras
to check whether the performance parameters meet the specifications quoted to
customers. NEMA has published several guidelines that describe the methods
to measure the performance parameters of gamma cameras and PET systems
[15.6, 15.10]. These guidelines provide standardized criteria for manufacturers to
measure and report the performance of their scanner. The IEC has also published
several technical reports [15.1715.19] describing the tests to be performed
during acceptance testing which reflects as closely as possible the clinical settings
in which gamma cameras and PET systems are operated. Most manufacturers
quote the performance of their systems according to NEMA guidelines [15.6].
The second level of testing is the acceptance testing performed after
the scanner arrives at the site. These tests should be performed by the user or
a third party, usually a qualified medical physicist, to determine whether the
578
system performs according to the manufacturers specifications and free of any

deficiencies, flaws or defects.
The baseline performance of the equipment will also be established. These
data provide guidance in the determination of the optimal operating parameters
for routine use and ensure that the imaging equipment meets regulatory
requirements for radiation safety [15.1].
These tests are usually very involved and require sophisticated phantoms
and dedicated software to calculate the performance parameters. Several national
and international agencies have set forth a range of tests, to be performed during
acceptance testing, that are easier (than the NEMA tests) to conduct. The American
Association of Physicists in Medicine (AAPM) is one of these agencies that has
produced several publications for testing gamma cameras during acceptance and
routine testing. The reports, AAPM 6, AAPM 9 and AAPM 22 [15.2015.22],
describe methods to perform acceptance testing on analogue, computer-aided
and SPECT capable gamma cameras, respectively. These reports describe tests
similar to those of NEMA. Moreover, the IAEA has published several books
describing the methods to perform tests on gamma cameras during acceptance,
reference and routine testing. Among them are TECDOC-317 and TECDOC-602
[15.16, 15.23]. The IAEA has recently published guidelines for QC and QA tests
for PET and PET/CT scanners [15.1].
During acceptance testing, the user should also conduct reference tests
which constitute the third level of testing. These tests reflect the performance
of the system under clinical settings, are easy to perform and can be performed
within an acceptable time frame. These tests, in addition to some other acceptance
tests, will establish the baseline performance characteristics for routine QC tests.
The results of routine tests are compared against the results of these tests.
Routine tests constitute the fourth level of testing. These are the tests
performed on a regular basis by users. Depending on the variability (in time) of
the performance parameter and its impact on image quality, test frequencies may
range from daily to annual. Several guidelines have been published on routine
tests, describing them and specifying their frequencies and the tolerance limits.
The IEC published standards 61675-2 and 61948-2 [15.18, 15.24] for gamma
camera routine testing including SPECT, and standard TR 61948-3 [15.25] for
PET routine testing. The AAPM has also published Report No. 52 [15.7] which
describes methods for measuring the quantification of SPECT performance.
Several issues regarding acceptance testing should be considered. Some
phantoms are used during acceptance testing and after major repair only, and
may be included in the purchasing contract. The manufacturer may also lend
their customers these phantoms during the period of acceptance testing. The slit
phantom used for testing linearity and intrinsic resolution of gamma cameras is a
typical example. The calculation of performance parameters from the image data
579
CHAPTER 15
in PET and the gamma cameras and SPECT systems may require sophisticated
software applications; thus, in such a case, the manufacturer must provide the
calculation software. The documentation for the acceptance test procedures
may be made available by the vendor. If needed, the recommendation of the
manufacturer should be followed, for instance, with regard to the amount of
activity required for each test. In multimodality imaging systems, additional
tests, which are not discussed in the existing guidelines, such as the accuracy of
image registration and attenuation correction, must also be conducted.
Before starting acceptance testing, the following additional issues should
be considered:
An accurate dose calibrator is an essential part of acceptance testing and
must, therefore, be available.
The required amount of radioactivity has to be arranged before starting
acceptance testing, so that the acceptance testing procedure does not
experience any interruption.
Proper calibration of the imaging system prior to acceptance testing is
of paramount importance. Any major erroneous calibration or lack of
calibration may result in an increase in commissioning cost and undue
delays in acceptance testing.
The order of the tests that will be conducted must be arranged so that any
malfunctioning or improper calibration can be discovered early on. This
will minimize the number of tests that must be repeated after recalibration
of the system.
If the medical physicist is not familiar with the system, a vendor
representative who knows how to operate the scanner and how to run the
calculation software should be present during acceptance testing.
All the required phantoms discussed in earlier sections of this chapter
should be made ready and prepared in advance.
15.4.2. Procurement and pre-purchase evaluations
When an institution decides to buy an imaging system, the administration
should start the planning properly by defining the purpose(s) for acquiring the
system and form a committee of a team of professionals to take on all of the
responsibilities from purchasing to setting up the system.
The purchasing committee should include the following professionals, as
defined in Ref. [15.1]:
Nuclear medicine and radiology physicians;
A medical physicist with experience in nuclear medicine;
580
If buying SPECT/CT or PET/CT, a medical physicist experienced in

diagnostic radiological physics should be included;
A medical physicist experienced in radiation therapy if the system will be
used in radiation therapy planning;
An administrator from the radiology department;
A radiation protection expert;
A person qualified in radiochemistry or radiopharmacy, if in-house
production of radiopharmaceuticals;
A nuclear medicine technologist;
A hospital management expert;
A bioengineering expert in imaging systems.
The role of this committee is to:
Choose the location;
Set the specifications of the system;
Prepare the tender documents;
Choose the proper system;
Supervise the installation process;
Supervise the acceptance and commissioning procedure.
This committee should start by choosing the proper space to host the
system. This location should be inside a radiation-controlled area, with the door
of the room opening to a closed vicinity (not to a public corridor). The room
should be wide enough to host the scanner, give accessibility to patient stretchers
and provide free space to maintenance engineers. If possible, the room should
be far away from MRI scanners to avoid any interference from their magnetic
field. If buying SPECT/CT and the CT sub-component will be used as a stand
alone system occasionally, it is advisable to have the scanner as close as possible
to the radiology CT scanner to act as a backup system when needed. This is also
true for the PET/CT systems if there is no on-site cyclotron. For scanners inside
institutions having a cyclotron, it is advisable to have the scanner as close as
possible to the cyclotron. This will allow the quick transfer of isotopes with very
short half-lives using dedicated lines or manual means.
The process of setting the specifications starts by agreeing on the purpose
for which the scanner will be used. Again, based on the applications that the
system will be used for, the different add-on components to be ordered will be
decided.
After defining all of the components, the specifications of the scanner and
each component should be set. To define a suitable specification, the committee
members should know what suitable systems are available that may meet their
581
CHAPTER 15
needs. After studying these systems, the required specifications should be set
with the aim of not excluding any available system initially. As a good practice,
one or several Excel work-sheets should be developed. The work-sheet(s)
should list all of the specifications, hardware, performance parameters, imaging
table, standard software, optional software, etc. Under each category, a list
of different specifications in that category should be listed with their limits.
Examples of hardware specifications are crystal(s) dimension and shape, number
of photomultiplier tubes, bore diameter and head movement ranges. Examples
of performance specifications are resolution, uniformity, dead time, SPECT
specifications, noise equivalent count rate and sensitivity. Examples of imaging
table specifications are pallet thickness, attenuation factor, scan range and speed,
minimum and maximum floor clearance, and weight limits. Knowing all of the
software that comes with the system on the acquisition and processing stations
and the optional ones available is necessary at this stage. The work-sheet(s) will
be distributed to all vendors as a soft copy, so that the answers from each will be
rearranged in one sheet to allow easy comparison of each specification between
vendors.
The tender should be prepared by the committee members and should
follow the institutions local regulations. It should include a summary of the
terms and conditions of the new equipment purchase deal. The following items
may be requested in a tender:
Name and model of the equipment.
Terms of pricing; way of payment, site preparation, accessories, etc.
Application specialist training.
System upgrade conditions.
Equipment references; short list of current users of similar system, local or
international.
Training of staff.
Equipment warranty.
Scheduling installation process and way of coordination.
Responsibility of site preparation, including removal of old equipment.
User and engineering manuals and equipment specifications (NEMA and
others).
Acceptance testing to be performed by a medical physicist (the system
should comply with NEMA or local specifications).
Commitments of the vendor to provide maintenance, and spare parts
readiness.
Specifications of local civil work and materials used.
582
Other steps that may assist the committee in the evaluation stage are:
Site visits: Manufacturers take the prospective customers to their reference
sites to evaluate the systems and listen to the users.
Evaluation of the clinical and phantom images provided by the
manufacturers: It is recommended that this be carried out on a common
imaging workstation for an objective comparison of different imaging
systems because each imaging workstation may process images differently
before displaying them on the screen. The medical physicist has to facilitate
the unbiased and blind comparison of the clinical images by the nuclear
medicine physicians.
Surveying centres with similar systems through a written questionnaire can
also be very effective and beneficial.
Inviting the vendor representatives to present their product in detail.
After thorough evaluation of all systems, the committee decides on the
most appropriate system upon considering the cost and other factors such as the
availability of a good maintenance service in the region.
After the system is chosen, the committee should supervise the installation
process. It should help the vendor representative to finalize all of the paper
work and get the access permits to the location. The system should be installed
completely with all the accessories and software ordered.
The local medical physicist or a private consultant should perform the
acceptance testing on the system. The committee should facilitate and make
available all of the necessary resources to the medical physicist to complete the
task and get the system ready for clinical use.
15.4.3. Acceptance testing as a baseline for regular quality assurance
As mentioned in Section 15.4.1, the medical physicist should produce
reference tests during acceptance testing. Tests should be acquired that are easy
to perform with less sophisticated procedures and that can be conducted within
an acceptable period by the user. These tests should reflect the performance of
the system in the working environment. The results of the routine tests should be
compared against the results of these reference tests.
For example, the medical physicist may acquire a five or ten million
counts uniformity image as a reference image for the system uniformity test
during the acceptance period. This is less sophisticated than the usual 30 million
counts uniformity image acquired for the acceptance testing. Another example is
acquiring a 10 million counts image for the bar phantom during the acceptance
testing and considering it a reference image. Some of the results of acceptance
583
CHAPTER 15
testing would be considered reference values to be used during routine testing.

The multiple window spatial registration, maximum count rate and system spatial
resolution values are examples of these tests.
15.4.4. What to do if the instrument fails acceptance testing
During acceptance testing, most of the performance parameters of the
system should be tested and compared with the manufacturers specifications.
These specifications should be required during tendering and be provided with
the system. If any of the test results do not meet the specifications, the analyses
should be re-evaluated carefully. Following this, the test should be repeated again,
paying close attention to any possible mistakes made during the acquisition and
processing of the data. The analysis should also be carried out carefully, making
sure that an accurate method has been followed.
If the problem persists, the engineer should be called to rectify the problem
and then it becomes the responsibility of the vendor to resolve the issue. The
engineer should look for any malfunction in the system, repair it and then
recalibrate the scanner. All of the required calibrations after this repair should
be performed and the system should be ready for testing. The medical physicist
should not start acceptance testing if the system still needs more calibration, as
some calibrations may readjust some parameters.
If two or three tests of the same parameter fail, the vendor should either
replace the affected part (if it was not done before) or replace the system. The
latter procedure should be the last option to be taken, as it will affect the routine
work of the clinic. The vendor should compensate the clinic and the medical
physicist (if a third party) for unnecessary delays.
15.4.5. Meeting the manufacturers specifications
The verification of performance specifications is one of the key reasons
for performing acceptance testing. Acceptance testing should follow the local
recommendations (in the institute or country) or one of the international bodies
recommendations. As was discussed earlier, for both PET and gamma cameras,
there are a number of guidelines set forth by various international bodies or
agencies (NEMA, IEC, AAPM and IAEA reports) as to what kind of tests should
be carried out.
Currently, there are task groups that have been formed by the AAPM
working on a new set of guidelines because the existing guidelines need
additional sets of performance tests to evaluate the hybrid systems as a whole,
and some modifications for the recently emerged new technologies are necessary.
584
The results of these tests should meet the specifications set by the
manufacturer, as they are usually one of the main reasons for selecting a particular
system. If one or more test results do not meet the manufacturers specifications,
the test should be repeated carefully. In the case of similar results, the vendor
engineer should rectify the problem at hand and then repeat the calibrations if
necessary.
REFERENCES
[15.1] INTERNATIONAL ATOMIC ENERGY AGENCY, Quality Assurance for PET and
PET/CT Systems, IAEA Human Health Series No. 1, IAEA, Vienna (2009).
[15.2] INTERNATIONAL ATOMIC ENERGY AGENCY, Quality Assurance for SPECT
Systems, IAEA Human Health Series No. 6, IAEA, Vienna (2009).
[15.3] INTERNATIONAL
COMMISSION
ON
RADIATION
UNITS
AND
MEASUREMENTS, Phantoms and Computational Models in Therapy Diagnosis and
Protection, ICRU Rep. 48, Bethesda, MD (1992).
COMMISSION
ON
RADIATION
UNITS
AND
MEASUREMENTS, Phantoms and Computational Models in Therapy Diagnosis and
Protection, ICRU Rep. 44, Bethesda, MD (1992).
[15.5] DEMIRKAYA, O., AL MAZROU, R., Performance test data analysis of scintillation
cameras, IEEE Trans. Nucl. Sci. 54 (2007) 15061515.
[15.6] NATIONAL ELECTRICAL MANUFACTURERS ASSOCIATION, Performance
Measurements of Gamma Cameras, Standards Publication NU 1-2007, NEMA (2007).
[15.7] GRAHAM, L.S., FAHEY, F.H., MADSEN, M.T., VAN ASWEGEN, A.,
YESTER, M.V., Quantitation of SPECT Performance: Report of Task Group 4,
Nuclear Medicine Committee (AAPM Report No. 52), Med. Phys. 22 4 (1995)
401409.
[15.8] NICHOLS, K.J., et al., Instrumentation quality assurance and performance, J. Nucl.
Cardiol. 13 (2006) 2541.
[15.9] HOFFMAN, E.J., CUTLER, P.D., DIGBY, W.M., MAZZIOTTA, J.C., 3-D phantom
to simulate cerebral blood flow and metabolic images for PET, IEEE Trans. Nucl.
Sci. 37 (1990) 616620.
[15.10] NATIONAL ELECTRICAL MANUFACTURERS ASSOCIATION, Performance
Measurements of Positron Emission Tomography, Standards Publication NU 2-2007,
NEMA (2007).
[15.11] BAILEY, D.L., JONES, T., SPINKS, T.J., A method for measuring the absolute
sensitivity of positron emission tomographic scanners, Eur. J. Nucl. Med. 18 (1991)
374379.
585
CHAPTER 15
[15.12] XU, X.G., CHAO, T.C., BOZKURT, A., VIP-Man: an image-based whole-body adult
male model constructed from color photographs of the visible human project for
multi-particle Monte Carlo calculations, Health Phys. 78 (2000) 476486.
[15.13] ZAIDI, H., XU, X.G., Computational anthropomorphic models of the human
anatomy: the path to realistic Monte Carlo modeling in radiological sciences, Annu.
Rev. Biomed. Eng. 9 (2007) 471500.
[15.14] CAON, M., Voxel-based computational models of real human anatomy: a review,
Radiat. Environ. Biophys. 42 (2004) 229235.
[15.15] ZUBAL, I.G., et al., Computerized three-dimensional segmented human anatomy,
Med. Phys. 21 (1994) 299302.
[15.16] INTERNATIONAL ATOMIC ENERGY AGENCY, Quality Control of Nuclear
Medicine Instruments, IAEA-TECDOC-602, IAEA, Vienna (1991).
[15.17] INTERNATIONAL ELECTROTECHNICAL COMMISSION, Medical Electrical
Equipment Characteristics and Test Conditions of Radionuclide Imaging Devices
Anger Type Gamma Cameras, 3rd edn, IEC 60789, IEC, Geneva (2005).
[15.18] INTERNATIONAL ELECTROTECHNICAL COMMISSION, Radionuclide Imaging
Devices Characteristics and Test Conditions Part 2: Single Photon Emission
Computed Tomographs, Edn 1.1, IEC 61675-2, IEC, Geneva (2005).
[15.19] INTERNATIONAL ELECTROTECHNICAL COMMISSION, Radionuclide Imaging
Devices Characteristics and Test Conditions Part 3: Gamma Camera Based
Whole Body Imaging Systems, 1st edn, IEC 61675-3, IEC, Geneva (1998).
[15.20] AMERICAN ASSOCIATION OF PHYSICISTS IN MEDICINE, Scintillation Camera
Acceptance Testing & Performance Evaluation, Report No. 6, AAPM, College Park,
MD (1980).
[15.21] AMERICAN ASSOCIATION OF PHYSICISTS IN MEDICINE, Computer-aided
Scintillation Camera Acceptance Testing, Report No. 9, AAPM, College Park,
MD (1982).
[15.22] AMERICAN ASSOCIATION OF PHYSICISTS IN MEDICINE, Rotating
Scintillation Camera SPECT Acceptance Testing and Quality Control, Report No. 22,
AAPM, College Park, MD (1987).
[15.23] INTERNATIONAL ATOMIC ENERGY AGENCY, Quality Control of Nuclear
Medicine Instruments, IAEA-TECDOC-317, IAEA, Vienna (1984).
Instrumentation Routine Tests Part 2: Scintillation Cameras and single photon
emission computed tomography imaging, IEC TR 61948-2, IEC, Geneva (2001).
Instrumentation Routine Tests Part 3: Positron Emission Tomographs, IEC TR
61948-3, IEC, Geneva (2005).
586
CHAPTER 16
FUNCTIONAL MEASUREMENTS
IN NUCLEAR MEDICINE
M.J. MYERS
Institute of Clinical Sciences,
Imperial College London,
London, United Kingdom
16.1. INTRODUCTION
The strength of nuclear medicine lies in using the tracer method to acquire
information about how an organ is or is not functioning as it should. This
modality, therefore, focuses on physiological organ function for diagnoses and
not on anatomical information such as X ray computed tomography (CT) or
magnetic resonance imaging.
The three aspects involved in the process are: (i) choice of radioactive
tracer, (ii) method of detection of the emissions from the tracer, and (iii) analysis
of the results of the detection. The radioactive tracers on which nuclear medicine
(or molecular imaging as it is increasingly being called) is based are designed
to participate in or trace a chosen function of the body. Their distribution is
then found by detecting and locating the emissions, usually photons, of the
radioactive tracer. The tracer may be involved in a metabolic process, such as
iodine in the thyroid, or it may take part in a physiological process because of its
physical make-up, such as macroaggregate of albumin (MAA) in the lungs.
A number of methods of detection can be used. One is imaging with a gamma
camera or positron emission tomography (PET) scanner in a number of modes:
static (showing an accumulated or integrated function), dynamic (showing the
variation of the function with time), whole body and tomographic (single photon
emission computed tomography (SPECT) and PET analysis). Another is simple
counting over areas of the body which can also be static or dynamic. Yet another
is through laboratory analysis of blood samples. Imaging often provides a rough
anatomical distribution of the function but, more importantly, a quantitative idea
of the extent of the function in the whole functional unit or in component parts
such as the right and left kidney. The anatomical picture has little of the detail
of the other modalities but may be a more direct tool in assessing pathology
since it provides primary information rather than displaying the anatomical
consequences of pathology such as changes in density. The images created can
587
CHAPTER 16
be directly related to the uptake of the radiopharmaceutical, either in terms of

counts or, with more sophisticated processing, with activity in becquerels. As
parametric images, they can also represent a parameter such as the distribution
of the ratio of ventilationV to perfusion Q, the V:Q ratio. The results can be
acquired and displayed as 2-, 3- or 4D images with time as the last dimension. In
addition, because of the unique property of nuclear medicine to be able to detect
specific radionuclides with different energy emissions, a number of functions
can be followed simultaneously. Thus, the ventilation of the lung traced by 81mKr,
a 190keV emitter, can be investigated at the same time as the perfusion of the
lung traced by the 140keV emitting 99mTc-MAA.
16.2. NON-IMAGING MEASUREMENTS
Non-imaging measurements refer to the analysis of data from radionuclide
procedures that are not derived from interpreting normal and pathological
patterns of uptake of tracer in images from gamma cameras and PET scanners.
Images may be used for non-imaging measurements but only to provide regions
of interest (ROIs) for subsequent quantification of function. A common example
of this is in the investigation of glomerular filtration in the kidney using a tracer
such as 51Cr-EDTA which can be measured from timed blood samples without
using images for information about morphological changes.
16.2.1. Renal function measurements
16.2.1.1. General discussion
Renal haemodynamic functions can be divided into measurements of
renal blood flow and glomerular filtration. The first depends on the supply of
blood to the cortical and extramedullary nephrons which are the functional
unit of the kidney. The second class of function depends on the transfer of
fluids across the glomerulus. A number of radioactive tracers may be used
depending on the function to be studied, the most common being 99mTc labelled
diethylenetriaminepentaacetic acid (DTPA), dimercaptosuccinic acid (DMSA)
and mercaptoacetyltriglycine (MAG3).
16.2.1.2. Glomerular filtration rate plasma clearance
Calculation of glomerular filtration rate (GFR) is used, for example, in
the general assessment of renal function and the monitoring of renal function
in patients undergoing therapy with nephritic drugs. Radioisotope measurements
588
FUNCTIONAL MEASUREMENTS IN NUCLEAR MEDICINE
depend on the assessment of plasma clearance with time as seen with blood
sampling of a tracer that is handled exclusively by glomerular filtration and does
not enter blood cells. The most common radiopharmaceutical used is 51Cr-EDTA,
though 99mTc-DTPA and 125I-iodothalamate are also seen.
GFR is obtained by constructing the clearance curve from one or a series of
timed measurements of plasma activity. In the multi-sample method, the expected
multi-exponential curve is defined accurately with samples taken at 10, 20 and
30min, and 2, 3 and 4 h, or approximated with samples taken at about 2 and 4h
or even at only one time point between 2 and 4 h. As taking multiple samples
over a period of hours may be impractical, further simplification of the process to
the taking of a single sample is attractive. An empirical relationship between the
apparent volume of distribution and the GFR has been derived and validated to
allow this at a less precise but acceptable accuracy.
The object of the measurements is to construct the total area under the
plasma clearance curve. It is sufficient for accuracy to assume a bi-exponential
curve with a fast and slow component between times of 10min and 4 h, ignoring
any initial very fast components. The zero time intercepts and rate constants for
the fast and slow components are C10 and , and C20 and , respectively. The
equation for GFR is:
Gfr =
injected activity
Q
Q0
(16.1)
= 0=
total area under plasma clearance curve A C10 + C20
where
injected activity has units of megabecquerels (MBq);
C10 and C20 are in count rates that are converted into megabecquerels per millilitre
(MBq/mL);
and and have units ofmin1, so that the GFR has units of millilitres perminute
(mL/min).
As the contribution to the whole area from the fast component is relatively
small and can be approximated without too much loss of accuracy, the equation
can be simplified to:
GFR =
Q0
(16.2)
C 20
589
CHAPTER 16
This produces an estimate of GFR that is obviously too small, though with
poor renal function the approximation has less of an effect. A correction factor
to convert the approximate GFR to the true GFR can be used. Although this
correction factor depends on the renal function, a figure of 1.15 can be used in
most cases.
GFR will vary with body size and is conventionally normalized to a
standard body surface area of 1.73 m2, though other normalization variables
such as the extracellular fluid volume have been suggested. The calculation for
GFR requires measurement of the activity injected into the patient as well as the
activity in the post-injection syringe, in a standard and in the blood samples. A
number of methods are available in practice. These are based on the difference
in weights of the pre- and post-injection syringe or on measurement of a fixed
volume in a dose calibrator or on known dilutions. The counts recorded by the
well counter measuring the small activities in the blood samples also have to be
calibrated in terms of megabecquerels per count rate (MBq/count rate).
16.2.1.3. Effective renal plasma flow measurements
Renal plasma flow, often referred to as renal blood flow, has been
investigated in the past using 131I or 123I labelled ortho-iodohippurate (hippuran)
or para-amino hippurate (PAH).
Hippuran is almost completely excreted by the renal tubules and extracted
on its first pass through the renal capillary system. As the extraction is not
quite 100%, the renal function measured is called the effective renal plasma
flow (ERPF). A modern variation is to use the 99mTc labelled tubular agent MAG3
as the 99mTc label is more available than 123I. However, the extraction fraction
of MAG3 at below 50% is inferior to that of hippuran, so the measurements of
ERPF are simply estimates.
The ERPF measurement is very much the same as that of GFR in that a
known activity of the radiopharmaceutical is injected and blood samples are
taken at intervals. The timing of the intervals is, however, earlier than for GFR
and occurs at 5min intervals and then at 30, 50, 70 and 90min. The resulting
two-exponential timeactivity curve is plotted from which the function is given
as:
ERPF =
C 10
Q0
+
C 20 (16.3)
using the symbols for GFR as above.
590
16.2.2.
14
C breath tests
The 14C urea breath test is used for detecting Helicobacter pylori infection
in cases of, for example, patients with duodenal and other ulcers following and
monitoring anti-H. pylori treatment. The test is based on the finding that the
bacterium H. pylori produces the enzyme urease in the stomach. As urease is not
normally found in that organ, its occurrence can, therefore, denote the existence
of H. pylori infection. The activity of 14C used in the test is very small, about
37 kBq, and the effective dose for this is also low, at less than 3 Sv.
To carry out the test, 14C urea is administered orally in the form of a
capsule. The urease in the stomach converts the urea to ammonia and 14C
carbon dioxide which is exhaled and can be detected in breath samples using a
liquid scintillation counter. The counter can measureminute quantities of the
emitting 14C. One or two samples are usually collected after 1030min using a
straw and balloon technique. Also counted are a known standard sample and a
representative background sample. Counting is done either locally or by sending
the samples to a specialized laboratory. The net disintegrations perminute (dpm)
registered by the counter are compared with standard values to assess the degree
of infection. dpm is given by:
dpm =
(S B)S t
(16.4)
(S t B)
where
S
B
is sample counts perminute;

is blank counts perminute;
and St is standard counts perminute.

A non-radioactive test employing 13C in place of 14C is also used.
Carbon-13 is measured by (non-radio)isotope ratio mass spectrometry. For 13C, a
baseline sample is required before the capsule is swallowed to compare with the
post-capsule sample.
16.3. IMAGING MEASUREMENTS
These include static image acquisition and analysis for quantitative
assessment of uptake, e.g. thyroid uptake measurement with organ delineation
and background subtraction. Furthermore, timeactivity curves can be derived
591
CHAPTER 16
from dynamic 2D imaging and quantitative parameters assessed from images,

e.g. renal function, gastric function, gall bladder emptying, gastrointestinal
transit and oesophageal transit. Timeactivity curves can also be derived from
dynamic 3D imaging with quantitative parameters assessed from physiologically
triggered images, such as cardiac ejection fraction measurement.
16.3.1. Thyroid
Measurement of thyroid function is one of the oldest techniques in nuclear
medicine, though the original radioisotope used, 131I, has been replaced by 99mTc
and 123I, and the collimated sodium iodide crystal uptake probe has almost
exclusively been replaced by the gamma camera.
Tests on the thyroid consist of both imaging the morphology of the organ and
assessing its uptake. Uptake consists of measuring the activity taken up by the
gland of an ingested or intravenously administered activity of radioactive iodine
or intravenously administered 99mTc-pertechnetate. The uptake mechanisms are
different for the two radioisotopes. Iodine is both trapped and organified by
thyroid follicular cells, a process more like the true thyroid function, whereas the
pertechnetate is simply trapped.
The tests allow assessment of functionality of thyroid lesions and
nodules, and investigations of thyroiditis and ectopic tissue. They can confirm
a diagnosis of an excess of circulating thyroid hormones, as in Graves disease
and toxic nodular goitre, and lead to a more quantitative approach to treatment of
hyperthyroidism with 131I.
Uptake of tumours secondary to thyroid cancer that have disseminated
through the body following surgery may also be assessed. In this case, whole
body images are made using 131I and a scanning gamma camera. The tumours can
be located and, in some cases, a quantitative measurement of uptake made, thus
allowing the effectiveness of treatment to be monitored.
The choice of radiopharmaceutical used can be dictated by the availability
and cost of the preferred isotope 123I and 99mTc. Each is suitable for use with a
gamma camera. Iodine-123 has a half-life of 13h and a emission of 160keV.
Technetium-99m has a half-life of 6h and an energy of 140keV. However,
the cyclotron produced 123I is not readily available and is expensive, while
99m
Tc-pertechnetate is readily available from the standard hospital molybdenum
generator. The timing of the uptake of the two isotopes is also different. While
iodine uptake is usually performed 1824h after injection or 26h after
ingestion, 99mTc is measured after 20min.
Uptake measurement with a scintillation probe involves counting over the
neck, over a thyroid phantom and over the thigh of the patient to simulate counts
in the neck that do not arise from the thyroid. The thyroid phantom (standardized
592
by, for example, the IAEA) consists of a small source of known activity in a
plastic cylinder offering the same attenuation as a neck. This acts as the standard.
Counts are obtained at a distance of about 25cm from the collimator face to
offset any inverse square errors due to different locations of the thyroid. The
percentage uptake U is then calculated from the formula:
U=
N T
100 (16.5)
Ca
where
N
T
is the counts in the neck;

is the counts in the thigh;
and Ca is the administered counts corrected for background.

The administered counts can be measured directly with an isotope calibrator
before the test or can be related to the activity in the thyroid phantom. Corrections
for decay are made throughout.
Similar measurements can be made with the gamma camera in place of the
probe, except that corrections for neck uptake can be made using ROIs over the
delineated thyroid and regions away from the gland. The camera can be fitted
with a pinhole collimator in order to provide a degree of magnification, although
the image is somewhat distorted and is subject to distance errors, in which case
a fixed distance (perhaps employing a spacer) is used. Images obtained with a
parallel-hole collimator may appear smaller but are not prone to distance effects,
though subject to the same attenuation. Anterior views are generally enough but a
lateral view may also be used to locate ectopic tissue.
Quantification of the uptake is achieved in two ways. By calibrating the
camera with a known activity in a suitable phantom, the activity injected into
the patient can be measured, or by measuring the injection directly in the syringe
before administration. Each will yield the sensitivity of the camera in terms of
counts per megabecquerel and allow the activities seen in the thyroid glands and
background to be calculated. The process is often an automatic one performed by
the camera computer software that delineates the outlines of the thyroid lobe(s)
and establishes a suitable background region used to correct for the presence
of activity in tissue overlying and underlying the thyroid tissue correction. It is
important that a local normal range is established and that the calibration of the
camera in terms of counts per megabecquerel is subject to a quality assurance
programme.
593
CHAPTER 16
A simple check on the quantification is to perform the test on a syringe

of activity as if the syringe is the thyroid gland, i.e. imaging the syringe as the
activity to be injected and re-imaging it as the activity taken up by the thyroid.
An uptake of 100% would be expected.
16.3.2. Renal function
The study of renal function has been a mainstay of nuclear medicine for
many decades and is an established efficient technique for, among other functions,
assessing renal perfusion, quantifying divided or individual kidney function, and
studying output efficiency and obstruction.
Two aspects of renal function are exploited: (i) glomerular filtration, i.e. the
transfer of fluids across the glomerulus, investigated by measuring the clearance
of 99mTc-DTPA (pentetate); and (ii) tubular secretion, investigated by measuring
the clearance of 99mTc-MAG3 (tiatide).
16.3.2.2. Renal function measurements
The basis of measurements of most renal functions is timeactivity curves
obtained by imaging the kidneys using a gamma camera. Views are obtained at
different times and over different periods after administration of one of a number
of radiotracers and often after some intervention. The result is usually a curve,
called a renogram, showing the rise and fall of counts in each kidney. This curve
is corrected for non-renal background counts using computer software, a feature
lacking in old counting methods using probes. Figure16.1 shows a series of
images obtained after administration of 99mTc-MAG3 (left), the ROIs drawn over
the two kidneys as well as the background region (top right), and the renogram
curves for the right kidney (RK) and left kidney (LK) (bottom right). The analysis
programmes, supported by commercial software providers, allow the calculation
of a number of renal function parameters, including relative perfusion, relative
function, mean andminimum transit times, and outflow efficiency.
There is an overabundance of different methods of analysis of the curves
to create a plethora of indices of function, some relying on the definition of
aspects of the renograms that have little basis in renal physiology. It is, therefore,
prudent to understand what is happening to the tracer as it traverses the kidneys,
and how it appears in the images and renogram. In fact, after adjustment for the
contributions of activity in the renal vasculature, the corrected curve displaying a
relatively fast rise and subsequent slower fall in activity can be described by two
distinct phases. The first spans the time of injection to the end of theminimum
594
transit time when the kidney contains the sum of all of the tracer extracted from
the blood and has, therefore, been termed the sum phase. The second starts at
the end of the first and reflects the net activity left after loss from the kidney and
has been called the difference phase. Other terms such as vascular spike and
secretory phase may not reflect purely renal function and are, therefore, not
very helpful.
FIG.16.1. Dynamic renal flow study after administration of 99mTc-MAG3 (left), regions of
interest drawn over the kidneys and background (top right), and corresponding renogram
curves for the right kidney (RK) and left kidney (LK) (bottom right).
What is helpful is often a quantitative comparison of the two kidneys with

derivation of a relative function. This may be calculated from Patlak plots or
from the uptake slope or the integral of the renogram curves. The programme
can calculate the relative perfusion and function from the retention functions.
The outflow curves for both kidneys and the value of the outflow at any selected
time can also be displayed. In the case of assessment of kidney transplants, other
aspects can be used to calculate relevant parameters.
A number of radioactive tracers may be used, depending on the function
to be studied. These functions are measured by 99mTc labelled DTPA, DMSA
and MAG3. Renal blood flow (or renal plasma flow) has been measured from
clearance of hippurate from the plasma.
595
CHAPTER 16
16.3.3. Lung function

The functions of the lung that are most investigated using nuclear medicine
techniques are regional ventilation and pulmonary blood flow or perfusion,
leading to studies of the ventilation perfusion ratio. Other studies cover
intrapulmonary vascular shunting, pulmonary granulocyte kinetics and lung
tissue permeability.
Lung air flow or ventilation imaging is carried out either with a gas such as
81m
Kr (a 13 s half-life radioisotope generated from 81Rb that has a 4.6h half-life)
or with an aerosol containing particles of sizes between 0.1 and 2m, typically
99m
Tc-DTPA or carbon particles (Technegas). The use of 133Xe gas has largely
been discontinued.
Lung blood flow or perfusion imaging is carried out with macroaggregates
or microspheres of denatured human serum albumin (MAA). These particles of
average size 2040m are larger than the lung capillaries and are trapped in the
capillary bed, distributing according to the blood flow to a region. Their main use
is to image pulmonary vascular disease, in particular, pulmonary embolism.
The two techniques are often employed together, either simultaneously
(e.g. 81Kr and 99mTc-MAA) or sequentially (99mTc aerosol and 99mTc-MAA). The
presence or absence of ventilation and/or perfusion is of clinical significance.
16.3.4. Gastric function
Nuclear medicine allows a full, non-invasive and quantitative assessment of
the way the oesophagus moves both solid and liquid meals to the stomach, how
the stomach handles these meals and how they transit through the gastrointestinal
tract.
16.3.4.1. Stomach emptying of solid and liquid meals
As the choice of both solid and liquid test meals determines the standard
values used as criteria for evaluating the function, a standard meal has been
agreed. Solid meals are based on preparations including eggs (into which 99mTc
sulphur colloid has been mixed), toast and water.
Anterior and posterior dynamic images are obtained at suitable intervals of
time following ingestion of the meal and are repeated for the same positioning
at hourly intervals for up to 4 h. The two views are obtained simultaneously or
sequentially, depending on the availability of a dual or single headed gamma
camera. The reason for the two view approach is to obtain a geometric mean
of the activity in the field of view that accounts for the movement of activity
596
between the anterior and posterior surfaces of the body. Relying on a simple
anterior view leads to artefacts due to differential attenuation of the 99mTc rays.
The data are analysed by drawing ROIs around the organs of interest
(stomach and parts of the gastrointestinal tract) and creating a decay corrected
timeactivity curve. An assessment of the gastric emptying function is made
from standard values. An alternative way of expressing the result is through the
half-emptying time.
16.3.4.2. Analysis of colonic transit
Colonic transport analysis can be performed using 111In labelled
non-absorbable material, such as DTPA or polystyrene micropellets administered
orally. Indium-111 is chosen rather than the more accessible 99mTc because of its
longer half-life (2.7 d) and the possibility of imaging over a longer time since
images are taken at, for example, 6, 24, 48 and 72 h.
As with the stomach procedures, a geometric mean parametric image of
anterior and posterior views may be used in the quantification, if this facility
is available in the nuclear medicine software. A geometric centre of the activity
(also called centre of mass) may be tracked over time by defining particular
segments in the colon, perhaps 511 in number (e.g the ascending, transverse,
descending, rectosigmoid and excreted stool), multiplying the individual segment
counts by weighting factors from 1 to 5, respectively, and summing the resulting
numbers. In addition to timeactivity curves for the individual segments, the
rate of movement of the geometric centre as a function of time can be assessed
by plotting this as a time position profile. A colonic half-clearance time may
be calculated and compared with historical normal control values of colonic
transport.
16.3.4.3. Oesophageal transit
This function is studied by imaging the transit of a bolus of radiolabelled
non-metabolized material such as 99mTc sulphur colloid. Either the whole
oesophagus may be included in an ROI and a timeactivity curve generated
for the whole organ, or a special display may be generated, whereby the counts
in successive regions of the oesophagus are displayed on a 2D spacetime
map called a condensed image. The counts in the regions are displayed in the
y direction as colour or grey scale intensities corresponding to the count rate
against time along the x axis (Fig.16.2). The result is a pictorial idea of the
movement of the bolus down the oesophagus.
597
CHAPTER 16
Bolus of
activity
travelling
down
oesophagus
imaged at
successive
times
Time (s)
Spacetime matrix
Distance
down
oesophagus
Time (s)
FIG.16.2. Oesophageal transit is imaged as a spacetime matrix. As the bolus of radioactivity
passes down the oesophagus, the counts from successive regions of interest, represented on a
grey scale, are placed in consecutive positions in the matrix in the appropriate time slot. A
normal transit will be shown as a movement of the bolus down and to the right in the matrix.
Retrograde peristalsis will be shown as a movement to the right and upwards.
16.3.4.4. Gall bladder ejection function

The gall bladder is investigated using hydroxy iminodiacetic acid labelled
with 99mTc. This is injected and can be imaged using a gamma camera after
excretion by the liver into the bile and as it passes through the gall bladder and
598
bile ducts. The gall bladder is then made to contract and empty by injecting a
hormone called cholecystokinin and the imaging of the gall bladder continued,
the whole test taking between 1 and 2 h. The amount of the radiolabel that
leaves the gall bladder is assessed by the difference in counts in the ROI over
the emptied gall bladder divided by the counts from the ROI over the full gall
bladder. Expressed as a percentage, this gives the ejection fraction. An ejection
fraction above 50% is considered as normal and an ejection fraction below about
35% as abnormal, suggesting, for example, chronic acalculous cholecystitis.
16.3.5. Cardiac function
The two main classes of cardiac function are blood flow in the myocardium
and in the blood pool and ventricles. Images are acquired in both planar and
tomographic modes, and the data may be acquired dynamically over sequential
time periods or as a gated study triggered by the electrocardiogram (ECG), or
as part of a first-pass study. The information is presented on a global or regional
basis as conventional or parametric images, or as curves from which quantitative
parameters may be derived. A range of pharmaceutical agents labelled with single
and positron emitting isotopes are used.
Cardiac functions that may be investigated with nuclear medicine techniques
run into many dozens, though relatively few are covered by standard clinical
practice and some are confined to research. A list of cardiac functions would,
therefore, include myocardial perfusion, myocardial metabolism of glucose and
fatty acids, myocardial receptors, left ventricular ejection fraction, first-pass
shunt analysis, wall motion and thickness, stroke volumes, cardiac output and its
fractionation, circulation times, systolic emptying rate, diastolic filling rate, time
to peak filling or emptying rate, and regional oxygen utilization.
Despite the usefulness of nuclear cardiology procedures, a worldwide
survey has shown a wide variation in their use and availability. There is a high
application rate in the United States of America (where cardiology accounts for
about half of nuclear medicine procedures) and Canada, less in western Europe
and Japan, and low application elsewhere such as in the Russian Federation,
Asia and some parts of South America. One reason for this pattern of use may
be the degree of access to training for physicians. Another reason may be that
gated SPECT imaging and analysis requires a high level of instrumentation and
software.
However, the procedures that can be carried out in any particular
department will depend very much on the nuclear medicine software provided by
the supplier of the gamma camera or from a specialized nuclear software supplier.
599
CHAPTER 16
A commercial suite of programmes will usually only offer a limited selection of

functional analysis. Typically, these include blood pool gated planar or SPECT
analysis for ventricular volumes and ejection fractions, and cardiac perfusion
analysis of gated SPECT images acquired under stress/rest conditions. The
sophistication of the programmes may also be limited in the extent of automation
of the preparation of cardiac images and their analysis. Whereas in the earlier
days of nuclear medicine a programmer could be asked to tailor programmes to
suit the need of the practitioner, and a high level language or macro-language
might be provided to allow this, the fashion nowadays is for a fixed set of
programmes dictated by the commercial supplier to satisfy a perceived general
need. A typical omission might, therefore, be list-mode data acquisition with the
consequent inability to trigger on selected parts of the ECG to omit ectopic heart
beats.
In addition to the aspect of the software provided is the choice of
radiopharmaceutical available. A wide choice is theoretically possible, depending
on the particular function to be explored. Thus, for example, myocardial
receptors can be investigated using 123I-MIBG (metaiodobenzylguanidine) or
11
C-hydroxyephedrine, myocardial glucose using 18F-FDG (fluorodeoxyglucose)
and fatty acid metabolism using 123I-heptadecanoic acid or BMIPP
(-methyl-p-iodophenylpentadecanoic acid). SPECT techniques use 201Tl,
99m
Tc-sestamibi and other perfusion agents. PET viability studies can employ
13
N-ammonia, 18F-FDG and 11C-acetate.
In terms of which gamma camera radiopharmaceutical is best to use,
there may be limitations in the availability or the licensing of a 201Tl or a 99mTc
product. As far as PET imaging goes, it is obvious that the non-research use of
short lived positron emitters such as 11C or 13N that would be useful in the study
of myocardial function would depend on the proximity of a cyclotron as well
as the funding to allow the production of the labelled products. It would also
presuppose the existence of an expensive PET/CT scanner for the procedures.
16.3.5.2. First-pass angiography
First-pass studies typically involve the acquisition of about 2000 frames of
data with a duration of about 50 ms following the bolus injection of autologous
red blood cells labelled in vivo or in vitro with 99mTc as they pass through the
right ventricle for the first time. A timeactivity curve derived from an ROI
over the ventricle shows a curve that rises to a peak and then falls off, the curve
also showing a saw tooth pattern corresponding to the filling and emptying of
the left ventricle during the cardiac cycle. The first-pass curve is illustrated in
Fig.16.3(a)alongside the volume curve in Fig.16.3(b), also used to derive the
ejection fraction obtained from a multiple-gated acquisition (MUGA) study.
600
Multiple-gated
count rate
EDV ESV
First-pass
count rate
Time (s)
EDV ESV = Ejection fraction

EDV
Time (s)
FIG.16.3. Curves obtained from (a) first-pass angiography and (b) a multiple-gated
acquisition study, showing how end diastolic volume (EDV) and end systolic volume (ESV),
used to determine ejection fraction, are represented in each technique.
By suitable positioning of the gamma camera in the right anterior oblique

view, this saw tooth can be used as an estimate of ejection fraction in the right
ventricle, a parameter that is only amenable to analysis in the first pass before
uptake in adjoining structures. The ejection fraction is derived from the ratio of
the peak of any saw tooth (the end diastolic volume (EDV))minus the value
of the next trough (the end systolic volume (ESV)) to the EDV. The ejection
fraction for the left ventricle would be assessed from the curve obtained by
viewing in the left anterior oblique position. Although this parameter is more
reliably obtained from a MUGA study, the first-pass procedure is much quicker
and may be suitable for patients who cannot tolerate the much longer MUGA
study. First-pass kinetics also provide a measure of left to right cardiac shunts
(where some activity goes from the left ventricle through a septal defect to the
right ventricle and, thus, recirculates between the heart and the lungs) and the
pulmonary systemic flow ratio (Qp:Qs), as well as systolic emptying and diastolic
filling rates and ventricular volumes.
16.3.5.3. The multiple-gated acquisition scan
The MUGA scan traces heart muscle activity from the distribution of the
administered radiopharmaceutical, allowing the calculation of the left ventricular
601
CHAPTER 16
ejection fraction and demonstrating myocardial wall motion. It may be obtained

while the patient is at rest and after physically or pharmacologically induced
stress.
The same radiopharmaceutical preparation is used as for the first-pass study
and a bolus is injected. The gamma camera views the patient in the left anterior
oblique position so as to best separate the projections of the two ventricles.
Dynamic images of the left ventricle in a beating heart are acquired at the same
time as the ECG and the results stored. A trigger (gating signal) corresponding
to the R-wave marks the start of each heart cycle and the start of each sequence of
images. The time period between successive R-waves (RR interval) is divided
into time intervals and the beat by beat left ventricular images corresponding
to each time interval are each integrated into a single combined gated image
that provides a stop motion image of the heart at intervals in its cycle. As any
one frame would not have enough data to provide sufficient counts and would,
therefore, be statistically unreliable, many frames at the same interval are
superimposed on each other. The signal for the start of each sequence is derived
from an ECG monitor connected to the patient that provides a short electronic
pulse as it detects the peak of an R-wave. Usually, about 32 equally time-spaced
frames (multiple gates) are used and these are defined between each RR
interval. Beats within 10% of the mean length are accepted. The result is a series
of images of the heart at end diastole and at end systole, and at stages in between
(Fig.16.4).
The image at end diastole when the heart has filled with blood contains
the maximum number of counts, and the end systolic image the least number.
A direct relationship is made between the number of counts in a region of the
ventricle and its volume.
For each frame, the computer, starting with an initial rough outline
provided by the operator, defines the boundary of the left ventricle. Depending
on the computer programme used, a different method of edge detection may be
employed. This may be based on an isocount contour or on a maximum slope
normal to the edge. As there is interference with the ventricular image from
activity seen in pulmonary blood, the computer will also define a suitable
background ROI close to the wall and correct the ventricular image at each stage.
The plot of the counts in the ventricular region against time forms the volume
curve that starts with the EDV, falls to the ESV level and rises again. The ejection
fraction is calculated as before from the standard formula:
EF =
602
EDV ESV
(16.6)
EDV
Simple differentiation of the volume curve provides indices of the

ventricular filling and emptying rates. Indeed, if the cardiac output is known, the
rates can be quantified in terms of millilitres perminute. Other information can
be obtained from phase/amplitude analysis of the sequence of images.
Beat by beat
images
Combined gated
images
FIG.16.4. The multiple-gated acquisition scan. The sequence of these gated images shows the
heart cycle with higher counts and better statistics than the individual images, allowing better
interpretation of the data than in Fig.16.3(b).
16.3.5.4. Myocardial perfusion imaging

Myocardial perfusion imaging allows a regional assessment of blood flow
to the heart and demonstrates areas of ischaemic myocardium where the blood
flow is diminished when the patient undergoes a stress test.
Imaging follows administration of specific radiopharmaceuticals such as
201
Tl-chloride, 99mTc-tetrofosmin or 99mTc-sestamibi after the patient has been
subjected to physical exercise or, if this is not suitable, to pharmacological
stress with vasodilators to raise the heart rate and stimulate the myocardium.
This induces a difference of uptake of radiolabel between normal and
ischaemic myocardial tissue that can be imaged and localized after planar
and/or ECG gated tomographic imaging. It is also possible to use positron
emitting radiopharmaceuticals such as the cyclotron produced 10min half-life
13
N-ammonia and generator produced 75 s half-life 82Rb, but there are the usual
limitations on applicability of PET, such as availability of the product or imaging
system.
Once a stable ECG pattern is observed, the patient is imaged using a
gamma camera operating in SPECT mode. ECG gating is applied throughout and
produces sets of 16 images at each acquisition angle. The stress images and their
603
CHAPTER 16
analysis may be compared with similar ones obtained with the patient at rest.
A number of protocols involving different times of examination and different
administrations of radioactivity have been devised to carry out the stress/rest
examinations in one rather than two days, given the potential long washout
periods involved. Imaging can be performed early (at about 15min) following
injection of 201Tl or 99mTc-sestamibi at rest or after the stress test and/or delayed
(after 14h or after 24 h) after injection at rest or under stress of the longer
lived 201Tl. These protocols give rise to different types of image. In general, the
imaging properties of 99mTc give superior images, though 201Tl is superior from a
physiological viewpoint as it is a better potassium analogue.
Conventional cardiac SPECT imaging may be carried out with a single or
double headed gamma camera using circular or elliptical orbits, the latter allowing
closer passes over the patient and, consequently, better resolution. Attenuation
correction may be performed on the emission images using an isotope or CT
X ray source. However, there is still debate as to the usefulness of attenuation
correction since this technique may give rise to artefacts due to mismatch of the
emission and transmission images on the fused images.
16.3.5.5. Technical aspects of SPECT and PET imaging
Thallium is not an ideal gamma camera imaging radionuclide, as it emits
rather low energy characteristic X rays between 69 and 80keV that are easily
attenuated (and, therefore, lost) in the body. The attenuation, by breasts and the
chest wall, varies for the different projections around the body and gives rise to
artefacts in the perfusion images if not corrected. The higher energy of 99mTc
(140keV) while still liable to attenuation, allows better collection of data from
the heart and less variation in the attenuation. Although SPECT/CT has still not
become a viable option (as has PET/CT where the two modalities have become
inseparable), this would be a better option for attenuation correction than the
rather cumbersome isotope attenuation correction devices that have been used in
the past.
Scattering of the photons before detection in the camera also leads to
problems in that their origin might be misplaced and loss from deeper structures
occurs. Recently, software to reduce the effects of scattering by modelling its
behaviour within the field of view has become available. Another source of
degradation of the image quality is the loss of resolution with distance from the
collimator face. Although high resolution collimators are usually chosen for
99m
Tc imaging, the basic resolution of the camera at the level of the heart is rather
poor. Again, software techniques to model this behaviour and correct for it have
become available.
604
Gamma camera images, unlike X ray ones, are always subject to lack of
counts and are, therefore, prone to statistical errors. Using a double headed rather
than a single headed system is, therefore, an advantage. There is still discussion
on the best angle between the heads and this may vary between less than 90 and
180. Scanning the patient with the collimator as close to the source of activity as
possible also ensures the best resolution, so a non-circular orbit is usually chosen.
Owing to the lack of accessible counts with 201Tl (its relatively long half-life
and high effective dose reducing the activity that can be administered), a general
purpose collimator is used, which is more efficient but less accurate than the high
resolution collimator used with 99mTc.
PET imaging based on the simultaneous detection of opposed annihilation
radiation from the original positron radiotracer is intrinsically tomographic
and invariably comes with anatomical land marking and attenuation correction
from the CT. It is also more sensitive and more accurate (because of its better
resolution) than SPECT and uptake of the radiopharmaceuticals can be
quantified absolutely. In theory, the use of 13N labelled ammonia and 18F-FDG
can differentiate more about the state of the myocardium, its blood flow and
metabolism, than the SPECT tracers.
Processing of the data starts with filtered back projection of the data using
standard filters such as a Butterworth or Hanning filter with appropriate cut-off
spatial frequencies and order, or an iterative reconstruction may be performed.
Attenuation correction may be applied. This stage produces a set of contiguous
slices parallel to the transverse axis of the patient which can be combined to
populate a 3D matrix of data. As the heart lies at an angle to this body axis,
a process of reorientation is performed. From the original matrix, the data that
lay parallel to the axes of the heart itself can be selected to form vertical long
axis (parallel to the long axis of the left ventricle and perpendicular to the
septum), horizontal long axis (parallel to the long axis of the left ventricle and
to the septum) and short axis (perpendicular to the long axis of the left ventricle)
slices through the myocardium of the particular patient, as shown in Figs16.5
and 14.12.
Cardiac processing software, working on features extracted from the shape
of the myocardium, allows easy automatic alignment which may also be operator
guided. The reoriented sections form three sets of images that are displayed in a
standard format to show, for example, the apex and heart surfaces at each stage of
gating of the heart cycle (Fig.16.4). The display may take several forms including
the simultaneous display of many sections in each of the three axes both at rest
and after stress, or as a moving image of the beating heart. The algorithms used
to carry out this process differ with the provider of the computer software and are
recognized under specific names.
605
CHAPTER 16
FIG.16.5. Principle of rearrangement of acquired axial tomographic images into slices

aligned with the axes of the heart.
Presentation of the slice data is often as a so-called polar diagram or

bulls-eye display illustrated in Fig.14.17. This allows the 3D information about
the myocardium, which would be difficult to interpret easily, to be depicted
as a simple, 2D, colour coded, semi-quantitative image. The process is often
described as imagining the myocardial surface as the peel of half an orange
which is flattened out to form the polar diagram. This is divided into accepted
segments and values, and colours associated with each segment. Again, there
is a variation of the exact form and mathematical basis of the polar diagram in
606
the commercial products available. This results in different looking maps that,
although individually validated, are not directly comparable. It would, therefore,
be prudent for one software package to be standardized at any one reporting
centre. The results from a particular study can be compared with a reference
image derived from a so-called normal database to allow a better estimation of
the extent of the defects. However, it is often difficult to match the population
being examined with the available validated normal data, for example, in terms
of gender and ethnicity.
BIBLIOGRAPHY
PETERS, A.M., MYERS, M.J., Physiological Measurements with Radionuclides in Clinical
Practice, Oxford University Press, Oxford (1998).
ZIESSMAN, H.A., OMALLEY, J.P., THRALL, J.H., Nuclear Medicine The Requisites,
3rd edn, Mosby-Elsevier (2006).
FURTHER READING
Recommended methods for investigating many of these functions
may be found on the web sites of the American Society of Nuclear
Medicine (www.snm.org), the British Nuclear Medicine Society
(www.bnms.org.uk) and the International Committee for Standardization in
Haematology (http://www.islh.org/web/published-standards.php).
607
CHAPTER 17
QUANTITATIVE NUCLEAR MEDICINE
J. OUYANG, G. EL FAKHRI
Massachusetts General Hospital
and Harvard Medical School,
Boston, United States of America
Planar imaging is still used in clinical practice although tomographic
imaging (single photon emission computed tomography (SPECT) and positron
emission tomography (PET)) is becoming more established. In this chapter,
quantitative methods for both imaging techniques are presented. Planar imaging
is limited to single photon. For both SPECT and PET, the focus is on the
quantitative methods that can be applied to reconstructed images.
17.1. PLANAR WHOLE BODY BIODISTRIBUTION MEASUREMENTS
Planar whole body imaging is almost always carried out by translating the
patient and bed in the z direction between opposed heads of a dual head standard
scintillation camera, typically in the anterior and posterior positions. The resulting
images are 2D projections of the 3D object being studied. The attenuation
of photons varies with the distance and the material the photons have to travel
through the object before reaching the detector. One approach to compensate for
attenuation in planar imaging is to perform conjugate counting with the geometric
mean, which consists of acquiring data from opposite views and combining them
into a single dataset. Figure17.1 shows an imaging object with uniform attenuation
viewed by two gamma detectors placed in opposite directions. A point source in
the object has attenuation depth d 1 and d 2 to detectors 1 and 2, respectively. The
projected counts P1 and P2 measured on detectors 1 and 2, respectively, are:
P1 = I 0 exp(d 1 ), P2 = I 0 exp(d 2 ) (17.1)
where
I 0
is the unattenuated number of counts that would be detected (in the
absence of attenuation);
and is the attenuation coefficient in the object.

608
These two conjugate views are combined using the geometric mean PG
defined as:
PG = P1P2 = I 0 exp(D / 2) (17.2)
where D is the total thickness and D= d 1 + d 2 .

The geometric mean depends on the total thickness, but not on the source
depths. This result is exact only for a point source. Corrections can be made for
extended sources [17.1]. The geometric mean using conjugate views is a popular
quantification method for planar imaging. The arithmetic mean (average of
opposite views) has also been proposed previously, but yields inferior results to
the geometric mean approach described here.
FIG.17.1. Illustration of conjugate viewing with the geometric mean for attenuation
compensation.
17.2. QUANTITATION IN EMISSION TOMOGRAPHY

17.2.1. Region of interest
Quantitative measurement of tracer uptake often requires the definition of
a region of interest (ROI), where the tracer uptake is to be quantified. A reliable
609
CHAPTER 17
although time consuming way to define an ROI is to draw the boundary of the
volume of interest on every slice containing the organ or tumour of interest. It is
difficult and time consuming to manually draw the boundary of an ROI because
the activity profile at the edge of the area of interest is not abrupt, but changes
slowly, and, therefore, deciding on a threshold to determine such a boundary is
not always straightforward. The manually drawn ROIs are, therefore, generally
not reproducible. Alternative semi-automatic and automatic methods use edge
detection techniques, which include count threshold, isocountours, maximum
slope or maximum count gradient, to improve reproducibility. Finally, another
approach that has been successfully used to determine organs or volumes with
a specific timeactivity behaviour in a dynamic acquisition is factor analysis of
dynamic sequences [17.2, 17.3].
17.2.2. Use of standard
When performing quantification from projections in the clinic, it can be
helpful to image a standard activity (known measured activity) along with the
patient (i.e. in the same projection). The standard (usually a small flask of the
radiotracer) provides a conversion between radioactivity concentration (MBq/mL)
and counts in the projections. It should be noted that the use of standards does not
guarantee accurate absolute quantification because the standard activity is not
affected by scatter, attenuation and partial volume effects in the same way as the
activity distribution in the patient.
17.2.3. Partial volume effect and the recovery coefficient
Ideally, the activity intensity within a region in a reconstructed image should
be proportional to the actual activity level in the region if scatter, attenuation,
randoms (PET only) and dead time corrections are properly applied, and if it is
assumed that there is very little noise. However, the partial volume effect (PVE)
significantly affects the quantification based on the size of the object of interest.
The PVE includes two different phenomena. One is the image blurring effect
caused by the finite spatial resolution. The blurring results in spillover between
regions. The image of a hot region, such as a tumour, appears larger and dimmer.
This limited resolution effect is theoretically described by a convolution operation.
The other PVE phenomenon is the so-called tissue fraction effect caused by the
fact that the boundaries of certain voxels do not match the underlying activity
distribution. The net PVE effect is the reduced contrast between the object and
the surrounding areas, as well as the reduced absolute uptake in a hot region. For
tumour imaging, the PVE can affect both the tumour apparent uptake and tumour
apparent size.
610
Recovery coefficient
The PVE is dependent on the size and shape of the region, the activity
distribution in the surrounding background, image spatial resolution, pixel size
and how uptake value is measured. The PVE correction is complicated by the
fact that not only activity from inside the region spills out but also activity from
outside the region spills in. As these two activity dependent flows are not usually
balanced, it is difficult to predict the overall PVE effect.
FWHM
FWHM
FWHM
FWHM
FWHM
Cylinder diameter (mm)

FIG.17.2. Recovery coefficient versus cylinder diameter for different full widths at half
maximum (FWHM) when imaging a cylinder filled with radioactivity in a cold background.
The ratio of the apparent concentration to true concentration is called

the recovery coefficient (RC). If the spatial resolution of the system can be
characterized at the location where the object of interest is (this is usually the
case in both SPECT and PET as the dependence of spatial resolution on location
is known), and if the size and shape of a region are known (i.e. from a different
modality such as computed tomography or magnetic resonance imaging), the RC
can be pre-calculated and then applied to the measured concentration value in
the region. Figure17.2 illustrates RC versus cylinder size for different SPECT
or PET spatial resolutions in the case of a cylindrical region which contains a
uniform level of radioactivity, surrounded by a cold (i.e. has no radiotracer
uptake) background. An RC close to one is obtained when the size of the object
is more than twice the full width at half maximum. As the RC depends on the
activity concentration in the region and in the surrounding background, the RC
will be different if the surrounding background is also hot (i.e. has radiotracer
611
CHAPTER 17
uptake). The RC correction method is a very simple method commonly used for
PVE correction in nuclear medicine [17.4]. However, the RC method can only be
used to correct the spillover between two structures. Geometric transfer matrix is
an approach that can account for the spillover among any number of structures
[17.5]. Deconvolution is another approach to perform PVE correction without
any assumption regarding tumour size, shape, homogeneity or background
activity. These three correction methods are used to correct the uptake value in a
region. It is also possible to model PVE in image reconstruction to obtain PVE
corrected images.
17.2.4. Quantitative assessment
Reconstructed images can be assessed qualitatively and quantitatively.
Qualitative interpretation is based on visualization that identifies regions with
abnormal patterns of uptake of the injected radiopharmaceutical as compared to
the known variants of radiotracer distribution. Quantitative assessment can be
either relative or absolute:
Target to background contrast: The target to background contrast is the ratio
between the concentration within the target region and the concentration
within the surrounding background. Therefore, contrast is considered a
relative quantification metric.
Radiotracer concentration: The radiotracer concentration (Bq/mL) is the
amount of radioactivity per unit volume within a defined ROI. Sometimes,
radiotracer concentration is converted into a different metric. For example,
the standardized uptake value (SUV) is the radiotracer concentration
normalized by injected dose and patient weight, and is mainly used to assess
tumour glucose utilization for fluorodeoxyglucose (FDG) PET. Therefore,
SUV is a semi-quantitative metric.
Kinetic parameters: A time sequence of PET measurements, i.e. dynamic
quantitative PET, makes it possible to measure tracer kinetics that describe
the interaction between the tracer and physiological processes. For example,
a water based tracer can be used to measure blood flow; a glucose based
tracer, such as FDG, can be used to measure metabolic rate. This is the
most accurate and absolute quantification metric that can be derived from
PET measurements. Usually, absolute quantification is best achieved with
PET because all projections are acquired simultaneously and, therefore,
dynamic imaging can be more easily performed than with rotating SPECT
cameras.
612
Clinical studies are generally analysed using qualitative and

semi-quantitative (e.g. contrast, SUV) information. This is also the case for
clinical trials and drug development.
17.2.4.1. Relative quantification using contrast ratio
Image contrast is the ratio of the signal level of a target relative to the signal
level in the surrounding background. The contrast ratio (CR) is defined as:
CR =
C T C B
(17.3)
CB
where C T and C b are the mean concentration within the defined target and
background region, respectively.
17.2.4.2. Relative quantification using the standardized uptake value
The SUV [17.6] is a widely used semi-quantitative metric to measure
tumour glucose utilization in, mostly, PET imaging. SUV (g/mL) is defined as:
SUV =
C i (kBq / mL)
(17.4)
A (kBq) / W (g)
where
Ci is either the mean (for SUVmean) or maximum (for SUVmax) decay-corrected
activity concentration (kBq/mL) within the defined region in the image (or
tissue);
A is the injected activity (kBq);
and W is the patient weight (g).
It is normally assumed that the density of tissue is equivalent to 1.0 g/mL,
such that the units effectively cancel and the SUV becomes a dimensionless
measure. The primary use of the SUV is to quantify activity in an ROI independent
of administered activity and patient weight. It has been shown that the SUV may
correlate with the metabolic rate of FDG in different tumour types, especially
when normalized for plasma glucose levels [17.7]. SUVmax instead of SUVmean
is often used because it is less sensitive to PVEs and it avoids including necrotic
or other non-tumour elements. However, SUVmax has lower reproducibility and a
larger bias than SUVmean as it is computed over a smaller number of voxels [17.8].
613
CHAPTER 17
To address this issue, a further term has been introduced, SUVpeak, which is
defined as the mean SUV value in a group of voxels surrounding the voxel with
the highest activity concentration in the tissue. The SUVs will be affected by the
level of image noise, which is affected by the reconstructed activity concentration,
the parameters chosen in the reconstruction algorithm and a myriad of other
factors. SUVpeak is intended to be a more robust measure than SUVmax.
The primary drawback with the SUV is that it is affected by many sources
of variability [17.9]. In addition to mathematical factors (e.g. ROI, noise and
PVE) that can alter the accuracy of the SUV, there are a number of biological
factors that can variably and unpredictably impact the SUV. Firstly, the SUV
calculation is based on the total administered dose. If a portion of the radiotracer
becomes interstitially infiltrated during intravenous injection, it is not routine
practice to correct for the activity that is trapped at the injection site and,
therefore, failing to circulate through the body. As a result, the calculated SUV
can be artificially low, because the total administered dose used in calculating
the SUV is greater than the actual dose reaching its intended intravascular target.
Secondly, the glucose avidity of tissues in the body is dependent on numerous
factors such as the presence of diabetes, insulin level and glucose level (the latter
two fluctuate widely depending on the patients most recent meal). If a patient
is diabetic, glucose is metabolized poorly by normal tissues, therefore leaving
more glucose available in the bloodstream to be metabolized by abnormally
glycolytic tissues such as tumour and infection, theoretically resulting in an
artificially elevated SUV. Conversely, if a diabetic patient has just received a
dose of insulin, the opposite effect may occur, lowering the SUV. In non-diabetic
patients, a patient who has recently eaten will have high glucose and insulin
levels, with a similar effect of lowering the SUV. Thirdly, FDG is cleared from
the body through urinary excretion. Patients with impaired renal function will
extract FDG more slowly from the bloodstream, leaving more FDG available
for metabolism by both normal and hypermetabolic tissues. Finally, body mass
is also a parameter used in calculating the SUV. In patients with a large number
of ascites, or other significant 3rd-spacing processes, the body mass of the
patient will be elevated by the presence of fluid that is neither intravascular nor
capable of uptake of radiotracer. Therefore, the denominator used in calculating
the SUV will be artificially large due to overestimation of the size of the patient,
theoretically causing an artificially low SUV to be calculated. Another factor
that plays a key role in standardization of tumour uptake is the reproducibility of
the measurement, and it has been previously shown that the correlation between
uptake measurements made in an identical manner at different clinical sites was
relatively low and significantly different to the standard reference measurement.
It is not uncommon that the SUV can vary 50% because of one or more such
effects. Therefore, the SUV should be properly corrected for all of these effects.
614
Another useful metric to assess tumour response to therapy is total lesion

glycolysis (TLG) defined as:
TLG = SUVmean V (17.5)
where V is the tumour volume (mL) that can be obtained using 3D contour
software.
TLG provides a measurement of the total FDG uptake in the tumour region,
which reflects total rather than average tumour metabolism.
17.2.4.3. Absolute quantification using kinetic modelling
Dynamic imaging consists of acquiring data as a series of time frames
that capture the timeactivity curve in each voxel over time, making it possible
to quantify tracer kinetics in vivo. Using a given radiotracer, the interaction of
the radiotracer with the bodys physiological, biochemical or pharmocokinetic
processes can be monitored. For example, glucose metabolism can be assessed
by FDG, a glucose analogue radiotracer. With an understanding of the underlying
physiological factors that control the tissue radioactivity levels, mathematical
models, known as kinetic models, can be constructed with one or more parameters
that describe the distribution of radiotracers as a function of time in the body
and fit the timeactivity curves in each voxel in the organ of interest. Kinetic
models used in nuclear medicine are based on compartments within a volume
or space within which the radiotracer becomes uniformly distributed almost
instantly, i.e. contains no significant concentration gradients. In other words,
compartmental modelling describes systems that vary in time but not in space.
More complicated kinetic models that include spatial gradients are generally not
applicable to nuclear medicine because of limited spatial resolution.
For a single-tissue compartmental model illustrated in Fig.17.3, the rate of
change in tracer concentration in a tissue is:
d C t (t )
= K 1C a (t ) k 2C t (17.6)
dt
where
Ct(t) is the tracer concentration in the tissue;
Ca(t) is the tracer concentration in the blood;
and K1 and k2 are the first order rate constants for the flux into the tissue and out
of the tissue, respectively.
615
CHAPTER 17
The solution for the above equation is:

C t (t ) = K 1C a (t ) exp(k 2t ) (17.7)
As the blood is the tracer delivery compartment, Ca(t) is also known as

the input function. The input function is directly measured from blood samples
or images of the blood, typically using the left ventricular blood pool. The
tracer concentration in the tissue Ct(t) is usually obtained from ROI analysis of
a dynamic series of images. Knowing both Ca(t) and Ct(t), regression analysis
can be applied to solve both K1 and k2. These kinetic parameters can be used to
interpret the underlying physiology. K1 is closely related to blood flow when the
extraction fraction is large, but is more related to permeability when the extraction
fraction is small. The ratio K1/k2 is the equilibrium volume of distribution that is
defined as Ct/Ca when the net tracer flux between compartments is zero after
sufficient time. By plotting the kinetic parameters as a function of the spatial
coordinates, parametric images of the radiotracer can be constructed.
Other applications of kinetic modelling in PET imaging include
measurement of glucose metabolic rate [17.10], and measurements of receptors
and neurotransmitters [17.11], etc.
FIG.17.3. Single-tissue compartmental model that describes the tracer exchange between
blood and tissue.
17.2.5. Estimation of activity

Bias, precision and accuracy are three important statistical concepts for
quantitative nuclear medicine. Bias is the difference between a population mean
of the measurements or test results and an accepted reference or true value.
616
The bias is mainly due to faulty measuring devices or procedures. One common
bias measure is defined as:
BIAS =
1
N
N
i=1
( x i t ) (17.8)
where
N
xi
is the total number of measurements;

is the measured value for the ith measurement;
and t is the true value.

Random error is a variable deviation and arises from the fluctuations in
experimental conditions, such as Poisson noise. Random error is also called
variance, but it is also often defined as the inverse, namely precision, referring to
the absence of random error. Precision can be quantified by the variance defined
as:
2 =
1
N
N
i=1
( x i x ) 2 (17.9)
where
x=
1
N
x
i=1 i
(17.10)
It is important to note that the calculation of precision does not require

knowledge of the true value. Therefore, precision alone cannot be used to
evaluate the performance of a measurement.
Bias and precision can be combined to assess the performance of a
measurement. Less biased and more precise measurements yield more accurate
estimations. Accuracy is, thus, defined as the overall difference between the
measured value and the true value. Accuracy can be quantified by the mean
square error (MSE) [17.12] defined as:
MSE =
1
N
N
i=1
( x i t ) 2 (17.11)
617
CHAPTER 17
It can be shown that [17.13]:

MSE = 2 + BIAS 2 (17.12)
To have the same scale as the mean value, precision is also quantified by
standard deviation, defined as the square root of variance. Similarly, accuracy is
also quantified as root MSE defined as the square root of MSE.
17.2.6. Evaluation of image quality
Image quality is a concept that has received a lot of attention recently
in an effort to better define what image quality is. A good surrogate for image
quality is image utility, i.e. the usefulness of an image for a particular detection or
quantification task, rather than measures of image properties, such as resolution,
contrast or stationarity of the point spread function, or of image fidelity, such
as normalized MSE. Measures of quantitative accuracy, precision and root MSE
are, of course, very useful when first assessing a new system or a quantification
method; however, for more rigorous evaluation or for definitive optimization
of data acquisition strategies, reconstruction techniques or image processing
procedures, it is recommended to carry out an objective assessment of image
quality based on detection or quantification tasks. Performance metrics for task
based estimation or detection tasks can be viewed as measures of image utility
which are the most clinically relevant bases on which to evaluate or optimize
imaging systems.
The most conclusive assessment of image quality is based on
human-observer studies. However, such studies are not routinely performed
clinically because they are time and resource consuming. Instead, a numerical
(or mathematical) observer is often used. It is beyond the scope of this chapter
to detail the different numerical observers used in SPECT and PET (for a review,
see Refs [17.13, 17.14]).
One of the simplest numerical observer methods is the non-prewhitening
matched technique, which is the optimal observer when images have uncorrelated
noise. Assuming N noise realizations of target-present image S and N noise
realizations of target-absent image B, the non-prewhitening signal to noise ratio
(SNRNPW) can be calculated as:
SNR NPW =
S T BT
1 / 2[ 2 ( S T )+ 2 ( BT )]
where T is the target matched filter.
618
(17.13)
More sophisticated numerical observers include the channelized Hotelling

observer that is a better surrogate for a human-observer under less ideal situations.
REFERENCES
[17.1] SORENSON, J.A., Quantitative measurement of radioactivity in vivo by whole-body
counting, Instrumentation of Nuclear Medicine (HINE, G.J., SORENSON, J.A.,
Eds), Academic Press, New York 2 (1974) 311348.
[17.2] BAZIN, J.P., DI PAOLA, R., GIBAUD, B., ROUGIER, P., TUBIANA, M., Factor
analysis of dynamic scintigraphic data as a modelling method. An application
to the detection of the metastases, Information Processing in Medical Imaging
(DI PAOLA, R., KAHN, E., Eds), INSERM, Paris (1980) 345366.
[17.3] HOUSTON, A.S., The effect of apex-finding errors on factor images obtained from
factor analysis and oblique transformation, Phys. Med. Biol. 29 (1984) 11091116.
[17.4] AVRIL, N., et al., Breast imaging with fluorine-18-FDG PET: quantitative image
analysis, J. Nucl. Med. 38 (1997) 11861191.
[17.5] ROUSSET, O.G., MA, Y., EVANS, A.C., Correction for partial volume effects in PET:
Principle and validation, J. Nucl. Med. 39(5) (1998) 904911.
[17.6] ZASADNY, K.R., WAHL, R.L., Standardized uptake values of normal tissues at PET
with 2-[fluorine-18]-fluoro-2-deoxy-D-glucose: variations with body weight and a
method for correction, Radiology 189 (1993) 847850.
[17.7] FLANAGAN, F.L., DEHADASHTI, F., SIEGEL, B.A., PET in breast cancer, Semin.
Nucl. Med. XXVIII (1998) 290302.
[17.8] KRAK, N.C., et al., Effects of ROI definition and reconstruction method on
quantitative outcome and applicability in a response, Eur. J. Nucl. Med. 32 (2005)
294301.
[17.9] KEYES, J.W., Jr., SUV: Standard uptake or silly useless value? J. Nucl. Med. 36
(1995) 18361839.
[17.10] PHELPS, M.E., et al., Tomographic measurement of local cerebral glucose metabolic
rate in humans with [18]-fluoro-2-deoxy-D-glucose: validation of method, Ann.
Neurol. 6 (1979) 371388.
[17.11] MINTUM, M.A., RAICHLE, M.E., KILBOURN, M.R., WOOTEN, G.F.,
WELCH, M.J., A quantitative model for the in vivo assessment of drug binding sites
with positron emission tomography, Ann. Neurol. 15 (1984) 217227.
[17.12] BURKHOLDER, D.L., Point estimation, International Encyclopedia of Statistics:
A Review (KRUSKAL, W.H., TANUR, J.M., Eds), J. Am. Stat. Assoc. 88 (1978)
251259.
[17.13] BARRETT, H.H., MYERS, K.J., Foundations of Image Science, John Wiley & Sons,
NJ (2004).
619
CHAPTER 17
[17.14] ABBEY, C.K., BARRETT, H.H., Human- and model-observer performance in

ramp-spectrum noise; effects of regularization and object variability, J. Opt. Soc. Am.
A Opt. Image Sci. Vis. 18 (2001) 473488.
620
CHAPTER 18
INTERNAL DOSIMETRY
C. HINDORF
Department of Radiation Physics,
Skne University Hospital,
Lund, Sweden
18.1. THE MEDICAL INTERNAL RADIATION DOSE FORMALISM
18.1.1. Basic concepts
The Committee on Medical Internal Radiation Dose (MIRD) is a
committee within the Society of Nuclear Medicine. The MIRD Committee was
formed in 1965 with the mission to standardize internal dosimetry calculations,
improve the published emission data for radionuclides and enhance the data on
pharmacokinetics for radiopharmaceuticals [18.1]. A unified approach to internal
dosimetry was published by the MIRD Committee in 1968, MIRD Pamphlet
No. 1 [18.2], which was updated several times thereafter. Currently, the most
well known version is the MIRD Primer from 1991 [18.3]. The latest publication
on the formalism was published in 2009 in MIRD Pamphlet No. 21 [18.4], which
provides a notation meant to bridge the differences in the formalism used by the
MIRD Committee and the International Commission on Radiological Protection
(ICRP) [18.5]. The formalism presented in MIRD Pamphlet No. 21 [18.4] will
be used here, although some references to the quantities and parameters used
in the MIRD primer [18.3] will be made. All symbols, quantities and units are
presented in Tables 18.1 and 18.2.
The MIRD formalism gives a framework for the calculation of the absorbed
dose to a certain region, called the target region, from activity in a source region.
The absorbed dose D is calculated as the product between the time-integrated
and the S value:
activity A
S (18.1)
D= A
The International System of Units unit of absorbed dose is the joule per kilogram
(J/kg), with the special name gray (Gy) (1 J/kg = 1 Gy).
621
CHAPTER 18
The time-integrated activity equals the number of decays that take place in a
certain source region, with units Bq s, while the S value denotes the absorbed
dose rate per unit activity, expressed in Gy (Bq s)1 or as a multiple thereof,
for example, in mGy (MBq s)1. The time-integrated activity was named the
cumulated activity in the MIRD Primer [18.3] and the absorbed dose rate per unit
activity was named the absorbed dose per cumulated activity (or the absorbed
dose per decay). A source or a target region can be any well defined volume,
for example, the whole body, an organ/tissue, a voxel, a cell or a subcellular
structure. The source region is denoted rS and the target region rT:
(r ) S(r r ) (18.2)
D(rT ) = A
S
T
S
The number of decays in the source region, denoted the time-integrated activity,
is calculated as the area under the curve that describes the activity as a function
of time in the source region after the administration of the radiopharmaceutical
(A(rS, t)). The activity in a region as a function of time is commonly determined
from consecutive quantitative imaging sessions, but it could also be assessed via
direct measurements of the activity on a tissue biopsy, a blood sample or via
single probe measurements of the activity in the whole body. Compartmental
modelling is a theoretical method that can be used to predict the activity in a
source region in which measurements are impossible.
(r ) =
A
S
A(r , t) dt (18.3)
S
The time-integration period TD, for which the time-integrated activity in the
source region is determined, is commonly chosen from the time of administration
of the radiopharmaceutical until infinite time, e.g. 0 to (Eq. (18.4)). However,
the integration period should be matched to the biological end point studied
in combination with the time period in which the relevant absorbed dose is
delivered.
(r ,T ) =
A
S D
TD
A(r , t) dt
S
(18.4)
The time-integrated activity coefficient a is defined as the time-integrated

activity divided by the administered activity A0, as can be seen in Eq. (18.5),
and has the unit of time. The time-integrated activity coefficient was named the
residence time in the MIRD Primer [18.3]. Figure 18.1 further demonstrates the
622
INTERNAL DOSIMETRY
concept. The area under the curve describing the activity as a function of time
TD
equals the area for the rectangle (
A(r , t) dt = a(r ) A
S
) and the time-integrated
activity coefficient can be described as an average time that the activity spends in
a source region.
a(rS ) =
(r )
A
S
(18.5)
A0
The S value is defined according to Eq. (18.6), which includes the energy emitted
E, the probability Y for radiation with energy E to be emitted, the absorbed
fraction f and the mass of the target region M(rT). The absorbed fraction
is defined as the fraction of the energy emitted from the source region that is
absorbed in the target region and equals a value between 0 and 1. The absorbed
fraction is dependent on the shape, size and mass of the source and target regions,
the distance and type of material between the source and the target regions, the
type of radiation emitted from the source and the energy of the radiation:
EY
(18.6)
M (rT )
Activity
S=
Time
FIG.18.1. The time-integrated activity coefficient (the residence time in the MIRD
Primer [18.3]) is calculated as the time-integrated activity divided by the injected activity,
which gives an average time the activity spends in the source region.
623
CHAPTER 18
The product of the energy emitted E and its probability to be emitted Y is denoted
, the mean energy emitted per decay of the radionuclide. The full formalism
also includes a summation over all of the transitions i per decay:
S(rT rS ) =
i (rT rS , i)
(18.7)
M (rT )
The absorbed fraction divided by the mass of the target region is named the
specific absorbed fraction :
(rT rS , E i ) =
(rT rS , E i )
(18.8)
M (rT )
The mass of both the source and target regions can vary in time, which means that
the absorbed fraction will change as a function of time after the administration,
and the full time dependent version of the internal dosimetry nomenclature
must be applied (Eq. (18.9)). This phenomenon has been noted in the clinic
for tumours, the thyroid and lymph nodes, and can significantly influence the
magnitude of the absorbed dose.
(rT rS , E i , t ) =
(rT rS , E i , t )
M (rT , t )
(18.9)
The total mean absorbed dose to the target region D(rT) is given by summing the
separate contributions from each source region rS (Eq. (18.10)). The self-absorbed
dose commonly gives the largest fractional contribution to the total absorbed
dose in a target region. The self-absorbed dose refers to when the source and
target regions are identical, while the cross-absorbed dose refers to the case in
which the source and the target regions are different from each other.
D(rT ) =
A (r )S(r
S
rS ) (18.10)
rS
The full time dependent version of the MIRD formalism can be found in
Eq. (18.11), where D denotes the absorbed dose rate:
TD
D(rT ,TD ) =
rS
624
TD
A(r , t)S(r
(r , t ) d t =
D
T
rS
rS , t ) dt (18.11)
INTERNAL DOSIMETRY
TABLE18.1. EXPLANATION OF SYMBOLS USED IN THE MEDICAL

INTERNAL RADIATION DOSE FORMALISM
Symbol
Parameter
Type of radiation
rS
Source region
rT
Target region
TD
Integration period
TABLE18.2. EXPLANATION OF THE SYMBOLS USED TO REPRESENT

QUANTITIES IN THE MEDICAL INTERNAL RADIATION DOSE
FORMALISM
Symbol
Quantity
Unit
(r ,T )
A
S D
Time-integrated activity
Bq s
a(rS ,TD )
Time-integrated activity coefficient
D(rT)
Absorbed dose to the target region rT
Gy

D
Absorbed dose rate
Gy/s
Mean energy of the ith transition per

nuclear transformation
J (Bq s)1 or MeV (Bq s)1
Ei
Mean energy of the ith transition
J or MeV
M(rT, t)
Mass of target region
kg
S(rT rS , t )
Absorbed dose rate per unit activity
mGy (MBq s)1
Time
Yi
Number of ith transitions per nuclear

transformation
(Bq s)1
(rT rS , E i , t )
Absorbed fraction
Dimensionless
(rT rS , E i , t )
Specific absorbed fraction
kg1
625
CHAPTER 18
18.1.2. The time-integrated activity in the source region

The physical meaning of the time-integrated activity in the source region
would be the number of decays in the source region during the relevant time
period. The time-integrated activity was named the cumulated activity in the
MIRD Primer [18.3].
The activity as a function of time A(t) can often be described by a
sum of exponential functions (Eq. (18.12)), where j denotes the number of
exponentials, Aj the initial activity for the jth exponential, the decay constant
for the radionuclide, j the biological decay constant and t the time after the
administration of the radiopharmaceutical. The sum of the j coefficients Aj
gives the total activity in the source region at the time of administration of the
radiopharmaceutical (t = 0):
A(rT , t ) =
A e
t ( + j )
(18.12)
The decay constant equals the natural logarithm of 2 (ln 2 = 0.693) divided by
the half-life. The decay constant in an exponential function matches the slope of
the curve it describes (in a linear-log plot of the function).
=
ln 2

T1/2
(18.13)
The physical half-life T1/2 and the biological half-life T1/2, j can be combined into
an effective half-life T1/2,eff according to Eq. (18.14). The effective half-life is
always shorter than both the biological and the physical half-lives alone.
1
1
1
(18.14)
=
+
T1/2,eff T1/2, j T1/2
The cumulated activity for the relevant time period is commonly calculated
as the time integral of an exponential function (Eq. (18.15)). However, other
functions could be used, with trapezoidal or Riemann integration (Fig. 18.2). The
trapezoidal and the Riemann methods could be reproduced with a higher accuracy
than the integration of an exponential, depending on how well the exponential fit
could be performed.
=
A
A(r ,0) e
S
626
t ( + j )
dt =
A(rS ,0)
(18.15)
+j
INTERNAL DOSIMETRY
(a)
(b)
(c)
12
24
36
48
60
Time
FIG.18.2. Demonstration of different methods to calculate the time-integrated activity using

a fit to an exponential function (a), trapezoidal integration (b) and Riemann integration (c).
Relevant biological data need to be acquired to perform absorbed dose

calculations with accuracy. The shape of the fitted curve, which describes the
activity as a function of time after the administration of the radiopharmaceutical,
can be strongly influenced by the number and timing of the individual activity
measurements (see Fig. 18.3). Three data points per exponential phase should be
considered the minimum data required to determine the pharmacokinetics, and
data points should be followed for at least two to three effective half-lives.
FIG.18.3. Two examples of the possible influences of curve fitting caused by the number
and the timing of activity measurements. The solid line gives the real activity versus time; the
dotted line represents the exponential curve fitted to the measurements, which are shown as
black dots.
The extrapolation from time zero to the first measurement of the activity in the
source region, and the extrapolation from the last measurement of the activity
in the source region to infinity, can also strongly influence the accuracy in the
time-integrated activity (see Fig. 18.4).
627
Activity
CHAPtER 18
Time
FIG. 18.4. Extrapolation before the first and after the last measurement point.
18.1.3. Absorbed dose rate per unit activity (S value)

The S value for a certain radionuclide and sourcetarget combination is
generated from Monte carlo simulations in a computer model of the anatomy.
The first models were analytical phantoms, in which the anatomy was
described by analytical equations. a coordinate system was introduced and
simple geometrical shapes such as spheres or cylinders were placed in the
coordinate system to represent important structures of the anatomy. several
analytical phantoms exist: adult man, non-pregnant woman, pregnant woman for
each trimester of pregnancy, children (from the newborn and up to 15 years of
age) as well as models of the brain, kidneys and unit density spheres.
Voxel based phantoms are the second generation of phantoms used for the
calculation of S values. These phantoms offer the possibility of more detailed
models of the anatomy. Voxel based phantoms can be based on the segmentation
of organs from tomographic image data, such as computed tomography (cT)
images.
The third generation of phantoms is created using non-uniform rational
b-spline (Nurbs). Nurbs is a mathematical model used in computer graphics
to represent surfaces. Nurbs provides a method to represent both geometrical
shapes and free forms with the same mathematical representation, and the
surfaces are flexible and can easily be rotated and translated. This means that
movements in time, such as breathing and the cardiac cycle, can be included,
allowing for 4-d representations of the phantoms [18.6].
anatomical phantoms for the calculation of S values for use in pre-clinical
studies on dogs, rats and mice have also been developed.
a common assumption in radionuclide dosimetry is that radiation
emissions can be divided into penetrating (p) or non-penetrating (np) and that the
absorbed fractions for these two types can be set to equal 0 and 1, respectively
(fp 0 and fnp 1). Electrons are often considered non-penetrating and photons
628
INTERNAL DOSIMETRY
as penetrating radiation, but this is an oversimplification. The validity of the

assumption is very dependent on the energy of the radiation in combination
with the size of the source region and must, therefore, be assessed on a case by
case basis, as is evident from Fig. 18.5. For electrons, the absorbed fraction is
greater than 0.9 if the mass of the unit density sphere is greater than 10 g and
the electron energy is lower than 1 MeV. This means that the approximation of
electrons as non-penetrating radiation is good at an organ level for humans, but
as the mass decreases, the approximation ceases to be valid. For photons, the
absorbed fraction is less than 0.1 if the mass of the sphere is less than 100 g and
if the photon energy is larger than 50 keV. The approximation of considering
photons as penetrating radiation is valid in most pre-clinical situations, but as the
mass increases, the approximation becomes inappropriate.
The self absorbed S values can be scaled by mass according to the following
equation:
S(rT rT ,scaled) S(rT rT ,tabulated)
M (rT ,tabulated)
M (rT ,scaled)
(18.16)
Absorbed fraction
Absorbed fraction
This is a useful method to adjust the S value found in a table to the true
weight of the target region. When scaling an S value, the absorbed fraction is
considered to be constant in the interval of scaling. The change in the S value is
then set equal to the change in mass of the target. It should be noted that linear
interpolation should never be performed in S value tables (Fig. 18.6).
FIG. 18.5. Absorbed fraction for unit density spheres as a function of the mass of the spheres
for mono-energetic photons (left) and electrons (right) (data from Ref. [18.7]).
A more sophisticated and probably more accurate way of recalculation of

the S value is by separating the total S value into two parts: one for penetrating
and one for non-penetrating radiation (Sp and Snp, respectively). If the absorbed
fraction for non-penetrating radiation is assumed to be equal to 1 (fnp = 1),
the S value for penetrating radiation can be calculated (Eqs (18.17)(18.19)).
629
CHAPTER 18
Ssphere (mGy/MBq-s)
The absorbed fractions for photons are relatively constant, so the S value for
penetrating radiation can be scaled by mass.
Mass (g)
FIG.18.6. Linear interpolation in S value tables gives S values that are too large. In this
particular case for a unit density sphere of 131I, linear scaling would give an S value that is
significantly greater than when scaling according to mass is performed.
S = S p + S np = S p +
S p = (S
np
m
np
m
m phantom
S recalculated = (S
mtrue
np
m
(18.17)
(18.18)
m phantom
mtrue
np
m
(18.19)
The absorbed fractions for photons and electrons vary according to the
initial energy and the volume/mass of the target region and, thus, the suitability
of the recalculation will also vary, as was discussed in the previous section
(Fig. 18.5).
The principle of reciprocity means that the S value is approximately the
same for a given combination of source and target regions, i.e. S(rTrS) is equal
to S(rSrT). The reciprocity principle is only truly valid under ideal conditions,
in regions with a uniformly distributed radionuclide within a material that is
either (i) infinite and homogenous or (ii) absorbs the radiation without scatter.
The ideal conditions are not present in the human body, although the reciprocity
principle can be seen in S value tables for human phantoms as the numbers are
almost mirrored along the diagonal axis of the table.
630
INTERNAL DOSIMETRY
S values for a sphere of a certain volume and material should be scaled

according to density if the material in the sphere is different from the material in
the phantom (Eq. (18.20)). The technique can be applied when an S value for a
unit density sphere is used for the calculation of the absorbed dose to a tumour
made up of bone or lung. However, it should be noted that an S value with the
correct mass could be chosen instead of scaling the S value for the correct volume
by the density.
S volume, material X = S volume, material Y
material Y
(18.20)
material X
18.1.4. Strengths and limitations inherent in the formalism

Two assumptions are automatically made when the MIRD formalism is
applied:
(a) The activity distribution in the source region is assumed to be uniform;
(b) The mean absorbed dose to the target region is calculated.
These assumptions are approximations of the reality. The strengths of the
MIRD implementation are its simplicity and ease of use. The limitation of these
assumptions is that the absorbed dose may vary throughout the region.
It is important to note that the MIRD formalism does not set any restrictions
on either the volume or the shape of the source or target, as long as uniformity
can be assumed. This means that the source and target volumes could be defined
so that the condition of uniformity is met.
The absorbed dose D is defined by the International Commission on
Radiation Units and Measurements as the quotient of the mean energy imparted
d and the mass dm [18.8]:
D=
d

dm
(18.21)
The absorbed dose is defined at a point, but it is determined from the mean
specific energy and is, thus, a mean value. This is more obvious from an older
definition of absorbed dose, where it is defined as the limit of the mean specific
energy as the mass approaches zero [18.9]:
D = lim z (18.22)
m 0
631
CHAPTER 18
The dosimetric quantity that considers stochastic effects and is, thus, not
based on mean values, is the specific energy z. The specific energy represents
a stochastic distribution of individual energy deposition events divided by the
mass m in which the energy was deposited [18.10]:
z=
(18.23)
The unit of the specific energy is joules per kilogram and its special name
is gray. Its relevance is especially important in microdosimetry which is the study
of energy deposition spectra within small volumes corresponding to the size of a
cell or cell nucleus.
The energy imparted to a given volume is the sum of all energy deposits i
in the volume:
=
(18.24)
The energy deposit is the fundamental quantity that can be used for the
definition of all other dosimetric quantities. Each energy deposit is the energy
deposited in a single interaction i:
i = in out + Q (18.25)
where
in is the kinetic energy of the incident ionizing particle;
out is the sum of the kinetic energies of all ionizing particles leaving the
interaction;
and Q is the change in the rest energies of the nucleus and of all of the particles
involved in the interaction.
If the rest energy decreases, Q has a positive value and if the rest energy
increases, it has a negative value. The unit of energy imparted and energy
deposited is joules or electronvolts. The summation of the energy deposits to
receive the energy imparted may be performed for one or more events, which is a
term denoting the energy imparted from statistically correlated particles, such as
a proton and its secondary electrons.
The absorbed dose is a macroscopic entity that corresponds to the mean
value of the specific energy per unit mass, but is defined at a point in space.
632
INTERNAL DOSIMETRY
When considering an extended volume such as an organ in the body, then for the
mean absorbed dose to be a true representation of the absorbed dose to the target
volume, either radiation equilibrium or charged particle equilibrium must exist.
Radiation equilibrium means that the energy entering the volume must equal
the energy leaving the volume for both charged and uncharged radiation. The
conditions under which radiation equilibrium are present in a volume containing
a distributed radioactive source are [18.11]:
The radioactive source must be uniformly distributed;
The atomic composition of the medium must be homogeneous;
The density of the medium must be homogeneous;
No electric or magnetic fields may disturb the paths of the charged particle.
Charged particle equilibrium always exists if radiation equilibrium exists.
However, charged particle equilibrium can exist even if the conditions for
radiation equilibrium are not fulfilled.
If only charged particles are emitted from the radioactive source (as is
the case for emitters such as 90Y and 32P), charged particle equilibrium exists
if radiative losses are negligible. Radiative losses increase with increasing
electron energy and with an increase in the atomic number of the medium.
The maximum energy for pure emitters commonly used in nuclear medicine
(e.g. 90Y, 32P and 89Sr) is less than 2.5 MeV and the ratio of the radiative stopping
power to the total stopping power is 0.018 and 0.028 for skeletal muscle and
cortical bone, respectively, for an electron energy of 2.5 MeV. This would imply
that the radiative losses can be neglected in internal dosimetry and charged
particle equilibrium can be assumed.
If both charged and uncharged particles (photons) are emitted (as is the case
with most radionuclides used in nuclear medicine), charged particle equilibrium
exists if the interaction of the uncharged particles within the volume is negligible.
A negligible number of interactions means that the photon absorbed fraction is
low. Photon absorbed fractions as a function of mass can be seen in Fig. 18.5, but
it should be pointed out that the relative photon contribution for a radionuclide
is also dependent on the energy and the probability of emission of electrons. For
example, the photon contribution to the absorbed dose cannot be disregarded for
111
In in a 10 g sphere, where the photons contribute 45% to the total S value.
18.1.4.1. Non-uniform activity distribution
The activity distribution is seldom completely uniform over the whole
tissue. This effect was theoretically investigated on a macroscopic level by
Howell et al. [18.12] by introducing activity distributions that varied as a
633
CHAPTER 18
function of the radius of a sphere. The non-uniformity in the activity distribution

can be overcome by redefining the source region into a smaller volume. This is
a feasible approach until the activity per unit volume becomes small enough to
cause a break-down of both radiation and charged particle equilibrium.
Redistribution of the radioactive atoms over time is responsible for
creating non-uniformities of the absorbed dose distribution over time. This
effect is handled indirectly in the MIRD formalism, which utilizes the concept
of cumulated activity, defined as the total number of decays during the time
of integration. However, in most practical applications of MIRD dosimetry,
heterogeneities of source distribution within organs are neglected.
18.1.4.2. Non-uniform absorbed dose distribution
If the activity of an or emitting radionuclide is uniformly distributed
within a sphere, then the absorbed dose distribution will be uniform from the
centre of the sphere out to a distance from the rim corresponding to the range of
the most energetic particle emission. If the radius of the sphere is large relative to
the particle emission ranges, then radiation equilibrium will be established except
at the rim and the mean absorbed dose will give a representative value of the
absorbed dose. If the radius of the sphere is of the same order as the range of the
emitted electrons, significant gradients in the absorbed dose distribution will be
formed at the borders of the sphere. As a rule of thumb, it can be assumed that
the absorbed dose at the border of the sphere will be half of the absorbed dose at
the centre. If the sphere is small compared to the range of the electrons, charged
particle equilibrium is never established and the absorbed dose distribution will
never be uniform inside the sphere. For emitting radionuclides, the absorbed
dose is uniform for almost all sized spheres, except within 7090 m from the
rim, corresponding to the particle range.
Interfaces between media, such as soft tissue/bone or soft tissue/air, will
cause non-uniformity in the absorbed dose distribution due to differences in
backscatter. This can be significant when estimating the contribution of absorbed
dose to the stem cells in the bone marrow from backscatter off the bone surfaces.
For planar geometry, the maximum increase in absorbed dose was 9%, as
determined by Monte Carlo simulations of 90Y. Experimental measurements with
32
P showed a maximal increase of 7%, in close agreement with the theoretical
estimates. For a spherical interface with a 0.5 mm radius of curvature, the
absorbed dose to the whole sphere showed a maximum increase for 0.5 MeV
electrons of as much as 12%.
Non-uniformities in the absorbed dose distribution will also be caused
by the cross-absorbed dose, i.e. when one organ is next to another such as lung
and heart. In human subjects, the separation between organs is sufficiently great
634
INTERNAL DOSIMETRY
that the cross-absorbed dose results from penetrating photon radiation only. It is
important to note that the cross-organ absorbed dose from high energy emitters,
such as 90Y and 32P, can be significant in preclinical small animal studies used to
study radiation toxicity. The importance of the cross-absorbed dose in comparison
to the self-absorbed dose strongly depends on both the S value and the relative
size of the time-integrated activity within the source and the target regions.
The MIRD formalism as such is equally applicable to any well defined
source and target region combinations [18.13, 18.14]. Depending on the volume
and dimensions of the regions, different types of emitted radiation will be of
different importance. To conclude the above discussion, a number of factors
causing non-uniformity in the absorbed dose distribution have been identified:
Edge effects due to lack of radiation equilibrium;
Lack of radiation equilibrium and charged particle equilibrium in the whole
volume (high energy electrons emitted in a small volume);
Few atoms in the volume, causing a lack of radiation equilibrium and
introduction of stochastic effects;
Temporal non-uniformity due to the kinetics of the radiopharmaceutical;
Gradients due to hot spots;
Interfaces between media causing backscatter;
Spatial non-uniformity in the activity distribution.
18.2. INTERNAL DOSIMETRY IN CLINICAL PRACTICE
Internal dosimetry is performed with different purposes, which would
require different levels of accuracy in the calculated absorbed dose, depending
on the subgroup:
Dosimetry for diagnostic procedures utilized in nuclear medicine;
Dosimetry for therapeutic procedures (radionuclide therapy);
Dosimetry in conjunction with accidental intake of radionuclides.
The dosimetry for a diagnostic procedure is performed to optimize the
procedure concerning radiation protection consistent with the requirements
of an accurate diagnostic test. This is an optimization of a clinical procedure
applicable to all persons. The most relevant would, therefore, be to utilize
the mean pharmacokinetics for the radiopharmaceutical for the calculation
of the time-integrated activity and S values based on a reference man
635
CHAPTER 18
phantom. The ICRP has published the absorbed dose per injected activity
for most radiopharmaceuticals used for diagnostic procedures in the clinic in
Publication 53 [18.5], with updates published in Publications 80 and 106 [18.15,
18.16].
The purpose of performing dosimetry for a patient that receives radionuclide
therapy is to optimize the treatment so as to achieve the highest possible absorbed
dose to the tumour, consistent with absorbed dose limiting toxicities. Thus,
individualized treatment planning should be performed that takes into account
the patient specific pharmacokinetics and biodistribution of the therapeutic agent.
The procedure to apply after an accidental intake of radionuclides must be
decided on a case by case basis. The procedure to apply will depend on the level
of activity, which radionuclide, the number of persons involved, whether the
dosimetry is performed retrospectively or as a precaution, and whether there is a
possibility to perform measurements after the intake.
18.2.2. Dosimetry on an organ level
Dosimetry on an organ level could be performed from activity quantification
using either 2-D or 3-D images. Two dimensional images may include whole
body scans or spot views covering the regions of interest. Three dimensional
single photon emission computed tomography (SPECT) is mostly a limited field
of view study that includes only the essential structures of interest. The advantage
of 3-D tomographic methods is that they avoid the problems associated with
corrections for activity in overlying and underlying tissues (e.g. muscle, gut
and bone), and corrections for activity in partly overlapping tissues (e.g. liver
and right kidney). Three dimensional positron emission tomography (PET) is
emerging as a powerful dosimetric tool because of the greater ease and accuracy
of radiotracer quantification with this modality.
S value tables for human phantoms can be found in MIRD Pamphlet No. 11
[18.17], in the OLINDA EXM software [18.18] and on the RADAR web site
(www.doseinfo-radar.com). OLINDA/EXM stands for organ level internal dose
assessment/exponential modelling, and is a software for the calculation of
absorbed dose to different organs in the body. OLINDA includes S values for
most radionuclides and for ten different human phantoms (adult and children at
different ages as well as pregnant and non-pregnant female phantoms). Tumours
are not included in the phantoms, although the S values for unit density spheres
provided in the software could be applied for the calculation of the self-absorbed
dose to the tumour. OLINDA also includes a module for biokinetic analysis,
allowing the user to fit an exponential equation to the data entered on the activity
in an organ at different time points. S values can be scaled by mass within
636
INTERNAL DOSIMETRY
OLINDA, thus allowing for a more patient specific dosimetry to be performed.

MIRDOSE [18.19] is the predecessor of OLINDA/EXM.
When calculating the absorbed dose with the MIRD formalism and using
tabulated S values for a phantom, for example, the reference man, it is assumed
that the patients anatomy is the same as that of the phantom. To employ the
MIRD scheme and yet make the dosimetry more patient specific, the S values
can be scaled to the mass of each patients target organ. Owing to the inverse
relation between the absorbed dose and the mass of the target region, scaling can
have a considerable influence on the result. The organ mass can be estimated
from CT, magnetic resonance imaging or ultrasound images, provided that the
anatomical size equals the functional size (the volume/mass of the organ that is
actually physiologically functioning and has an activity uptake).
S patient S phantom
m phantom
m patient
(18.26)
Since it requires a great deal of work to determine the mass of every organ
for each patient, it was suggested that the S values might be scaled to the total
mass of the patient. This is a more crude method, assuming that the organ size
follows the total mass of the body. The lean body weight of the patient should be
used to avoid unrealistic values of the organ mass and, thus, the S values due to
obese or very lean patients.
S patient S phantom
m TB,phantom
m TB,patient
(18.27)
Tumours are not included in reference man phantoms. However, S values

could be used for spheres of the correct mass to get an approximation of the
self-absorbed dose to the tumour. The drawback with this method is that neither
the contribution from the cross-absorbed dose from activity in normal organs to
the tumour nor the cross-absorbed dose from activity in the tumour to normal
organs can be included in the calculations.
18.2.3. Dosimetry on a voxel level
The activity in an image could be quantified on a voxel level, to display
the activity present in each voxel. Images that display the activity distribution
at different points in time after injection may be co-registered to each other to
allow for an exponential fit on a voxel by voxel basis. A parametric image that
gives the time-integrated activity (the total number of decays) on a voxel level
637
CHAPTER 18
F(r/r0)
can, thus, be calculated. Parametric images that display the biological half-life
for each voxel could also be produced by this technique.
The registration of the images acquired at different points in time after the
administration becomes essential for the accuracy that can be achieved in the
calculation of the time-integrated activity on a voxel level. Another important
factor that determines the accuracy in the time-integrated activity and, thus, in
the absorbed dose, is the acquired number of counts per voxel (a random error),
the accuracy in the attenuation correction (systematic error) and the calibration
factor that translates the number of counts to the activity (random and systematic
errors). Multimodality imaging such as SPECT/CT and PET/CT facilitates the
interpretation of the images as the CT will provide anatomical landmarks to support
the functional images, which could change from one acquisition to the next.
A dose point kernel describes the deposited energy as a function of distance
from the site of emission of the radiation. Figure 18.7 displays a dose point
kernel for 1 MeV mono-energetic electrons. Convolution of a dose point kernel
and the activity distribution from an image acquired at a certain time after the
injection gives the absorbed dose rate. Dose point kernels provide a tool for fast
calculation of the absorbed dose on a voxel level. However, the main drawback
is that a dose point kernel is only valid in a homogenous medium, where it is
commonly assumed that the body is uniformly unit density soft tissue.
Monte Carlo simulations that use the activity distribution from a functional
image (PET or SPECT) and the density distribution from a CT image avoid the
problem of non-uniform media, although full Monte Carlo simulations are time
consuming. EGS (electron gamma shower), MCNP (Monte Carlo N-particle
transport code), Geant and Penelope are commonly used Monte Carlo codes.
r/r0
FIG.18.7. A scaled dose point kernel for 1 MeV electrons [18.20]. r/r0 expresses the
distance scaled to the continuous slowing down approximation range of the electron and
F (r / r , E ) d(r / r ) = 1.
0
638
INTERNAL DOSIMETRY
The concept of dosevolume histograms (DVHs), extensively used to

describe the tumour and organ dose distribution in external beam radiotherapy,
can be used to display the non-uniformity in the absorbed dose distribution from
radionuclide procedures. A differential DVH shows the fraction of the volume
that has received a certain absorbed dose as a function of the absorbed dose,
while a cumulative DVH shows the fraction of the volume that has received an
absorbed dose less than the figure given on the x axis. A truly uniform absorbed
dose distribution would produce a differential DVH that shows a single sharp
( function) peak and a step function on a cumulative DVH. Since the mean
absorbed dose in internal dosimetry may be a poor representation of the absorbed
dose to the tissue, as discussed above, the use of DVHs might be used to assist
the correlation between absorbed dose and biological effect.
REFERENCES
[18.1] STELSON, A.T., WATSON, E.E., CLOUTIER, R.J., A history of medical internal
dosimetry, Health Phys. 69 (1995) 766782.
[18.2] LOEVINGER, R., BERMAN, R.M., A schema for absorbed-dose calculations for
biologically-distributed radionuclides, MIRD Pamphlet No. 1, J. Nucl. Med. 9
Suppl. 1 (1968) 714.
[18.3] LOEVINGER, R., BUDINGER, T.F., WATSON, E.E., MIRD Primer for Absorbed
Dose Calculations (Revised Edition), The Society of Nuclear Medicine, MIRD,
Reston, VA (1991).
[18.4] BOLCH, W.E., ECKERMAN, E.F., SGOUROS, G., THOMAS, S.R., A generalized
schema for radiopharmaceutical dosimetry standardization of nomenclature, MIRD
Pamphlet No. 21, J. Nucl. Med. 50 (2009) 477484.
Radiation Dose to Patients from Radiopharmaceuticals, Publication 53, Pergamon
Press, Oxford (1987).
[18.6] SEGARS, W.P., TSUI, B.M., FREY, E.C., JOHNSON, G.A., BERR, S.S.,
Development of a 4-D digital mouse phantom for molecular imaging research, Mol.
Imaging Biol. 6 (2004) 149159.
[18.7] STABIN, M.G., KONIJNENBERG, M.W., Re-evaluation of absorbed fractions for
photons and electrons in spheres of various sizes, J. Nucl. Med. 41 (2000) 149160.
COMMISSION
ON
RADIATION
UNITS
AND
MEASUREMENTS, Fundamental Quantities and Units for Ionizing Radiation,
Rep. 60, ICRU, Bethesda, MD (1998).
COMMISSION
ON
RADIATION
UNITS
AND
MEASUREMENTS, Radiation Quantities and Units, Rep. 33, ICRU, Bethesda, MD
(1983).
639
CHAPTER 18
COMMISSION
ON
RADIATION
UNITS
MEASUREMENTS, Microdosimetry, Rep. 36, ICRU, Bethesda, MD (1983).
AND
[18.11] ATTIX, F.H., Introduction to Radiological Physics and Radiation Dosimetry, John
Wiley & Sons, New York (1986).
[18.12] HOWELL, R.W., RAO, D.V., SASTRY, K.S.R., Macroscopic dosimetry for
radioimmunotherapy: Nonuniform activity distributions in solid tumours, Med.
Phys. 16 (1989) 6674.
[18.13] HOWELL, R.W., The MIRD schema: From organ to cellular dimensions, J. Nucl.
Med. 35 (1994) 531533.
[18.14] KASSIS, I.E., The MIRD approach: Remembering the limitations, J. Nucl. Med. 33
(1992) 781782.
Radiation Dose to Patients from Radiopharmaceuticals (Addendum to ICRP
Publication 53), Publication 80, Pergamon Press, Oxford and New York (1998).
Radiation Dose to Patients from Radiopharmaceuticals (Addendum 3 to ICRP
Publication 53), Publication 106, Elsevier (2008).
[18.17] SNYDER, W.S., FORD, M.R., WARNER, G.G., WATSON, S.B., MIRD Pamphlet
No. 11, S, Absorbed Dose per Unit Cumulated Activity for Selected Radionuclides and
Organs, The Society of Nuclear Medicine, Reston, VA (1975).
[18.18] STABIN, M.G., SPARKS, R.B., CROWE, E., OLINDA/EXM: The second-generation
personal computer software for internal dose assessment in nuclear medicine, J. Nucl.
Med. 46 (2005) 10231027.
[18.19] STABIN, M.G., MIRDOSE: Personal computer software for internal dose assessment
in nuclear medicine, J. Nucl. Med. 37 (1996) 538546.
[18.20] BERGER, M., Improved point kernels for electron and beta-ray dosimetry,
NBSIR 73107, National Bureau of Standards (1973).
640
CHAPTER 19
RADIONUCLIDE THERAPY
G. FLUX1, YONG DU2
Joint Department of Physics1 and Nuclear Medicine2,
Royal Marsden Hospital and Institute of Cancer Research,
Surrey, United Kingdom
19.1. INTRODUCTION
Cancer has been treated with radiopharmaceuticals since the 1940s. The
radionuclides originally used, including 131I and 32P, are still in use. The role of
the physicist in radionuclide therapy encompasses radiation protection, imaging
and dosimetry. Radiation protection is of particular importance given the high
activities of the unsealed sources that are often administered, and must take into
account medical staff, comforters and carers, and, as patients are discharged
while still retaining activity, members of the public. Regulations concerning
acceptable levels of exposure vary from country to country. If the administered
radiopharmaceutical is a emitter, then imaging can be performed which may be
either qualitative or quantitative. While a regular system of quality control must
be in place to prevent misinterpretation of image data, qualitative imaging does
not usually rely on the image corrections necessary to determine the absolute
levels of activity that are localized in the patient. Accurate quantitative imaging is
dependent on these corrections and can permit the distribution of absorbed doses
delivered to the patient to be determined with sufficient accuracy to be clinically
beneficial.
Historically, the majority of radionuclide therapies have entailed the
administration of activities that are either fixed, or may be based on patient
weight or body surface area. This follows methods of administration necessarily
adopted for chemotherapy. However, given that in vivo imaging is possible for
many radiopharmaceuticals and that the mechanism of therapy is the delivery
of a radiation absorbed dose, the principles of external beam radiation therapy
apply equally to radionuclide therapies. These are summarized in European
Directive 97/43:
For all medical exposure of individuals for radiotherapeutic purposes
exposures of target volumes shall be individually planned; taking into
account that doses of non-target volumes and tissues shall be as low as
641
CHAPTER 19
reasonably achievable and consistent with the intended radiotherapeutic

purpose of the exposure.
In this directive, the term radiotherapeutic specifically includes nuclear
medicine for therapy.
Radionuclide therapy is a rapidly expanding cancer treatment modality,
both in terms of the number and range of procedures given, and many new
radiopharmaceuticals are now entering the market. At present, there is a paucity
of guidelines governing levels of activity to administer and these vary widely
according to local protocols. The application of internal dosimetry to therapeutic
procedures will allow the data to be collected on which to establish the evidence
necessary to optimize treatment protocols.
For many therapy procedures, dosimetry studies have been conducted.
These have demonstrated that a wide range of absorbed doses are delivered both
to target tissues and to normal tissues from the administration of fixed activities
due to variations in uptake and retention of a radiopharmaceutical. It is likely
that, in conjunction with patient variations in radiosensitivity, this accounts for
the variable response seen with radionuclide therapy.
Recent advances in the quantification of single photon emission computed
tomography and positron emission tomography data, and increased research into
patient specific rather than model based dosimetry, have led to the possibility of
personalizing patient treatments according to individual biokinetics.
19.2. THYROID THERAPIES
19.2.1. Benign thyroid disease
Benign thyroid disease (typically hyperthyroidism or thyrotoxicosis) is
most commonly caused by Graves disease, an autoimmune disease affecting
the whole thyroid gland and causing it to swell. Thyroid toxic nodules,
consisting of abnormal thyroid tissue, are also responsible for overactive thyroid
glands. Iodine-131 NaI (radioiodine) has been used since the 1940s to treat
hyperthyroidism successfully.
There is a long standing wide acceptance of radioiodine as a treatment for
hyperthyroidism, particularly for patients with solitary toxic adenoma, although
treatment protocols vary, and there is limited evidence to compare long term
results from surgery, anti-thyroid drugs or radioiodine. Guidelines are available
from the European Association of Nuclear Medicine (EANM) and the American
Thyroid Association, as well as from individual countries including Germany and
the United Kingdom.
642
19.2.1.1. Treatment specific issues

Standard administrations can vary from 200 to 800MBq, depending on the
patient situation and local practice. While persistence of symptoms will result
from inadequate treatment, entailing further administrations, it can be argued
that patients should not receive more activity than is necessary to render them
euthyroid. Therefore, in common with other radionuclide therapies, a major
issue is that of personalized treatment based on patient specific dosimetry. This
necessitates determination of the thyroid volume and calculation of the activity
required to administer a fixed absorbed dose based on a tracer study. A range
of methods have been followed to determine the thyroid volume, using, for
example, ultrasound, 123I-NaI or 124I-NaI, and there is some discrepancy in the
reported absorbed doses required to achieve euthyroidism. Further research work
would certainly benefit patients.
Radiation protection advice must be given to a patient undergoing
radioiodine treatment although standard treatments can usually be conducted on
an out-patient basis.
19.2.2. Thyroid cancer
Thyroid cancer accounts for less than 0.5% of all cancers and there
are 28000 new cases diagnosed each year in Europe and the United States
of America. Papillary and follicular thyroid cancer account for 8090% of
cases, with the remainder being anaplastic carcinomas, medullary carcinomas,
lymphomas and rare tumours. Increased risk is associated with benign thyroid
disease, radiation therapy to the neck and poor diet. As with benign thyroid
disease, thyroid cancer has also been treated with radioiodine for over sixty
years and in conjunction with total or near total thyroidectomy is widely used
for an initial ablation of residual thyroid tissue. Up to 20% of cases may present
with metastatic disease, usually to the lungs or bones although also to liver and
brain. Treatment for distant metastases usually involves further and often higher
administrations of radioiodine. This treatment is the most common application
of radionuclides for therapy and is very successful, with complete response rates
of 8090%. Nevertheless, the disease can prove fatal in a higher proportion of
patients that are most at risk, which include the young and the elderly.
There are a number of controversies concerning the treatment of thyroid
cancer with radioiodine. These include the extent of a low iodine diet prior to
administration, levels of activity to administer for ablation or for therapy, and the
643
CHAPTER 19
time interval between ablation and the determination of success, which itself is
subject to debate. The issue that most affects the physicist is that of standardized
versus personalized treatments, which has been debated since the early 1960s.
Fixed activities given for ablation can vary from 1100 to 4500MBq, and
those given for subsequent therapy procedures can be in excess of 9000MBq.
Published guidelines report the variation in fixed activities but do not make
recommendations concerning these levels.
It has been conclusively demonstrated in a number of dosimetry studies
that patients administered fixed activities of radioiodine receive absorbed doses
to remnant tissue, residual disease and to normal organs that can vary by several
orders of magnitude. This potentially has important consequences, as it implies
that, in many cases, patients may be receiving less absorbed dose than is required
for a successful ablation or therapy, while in other cases patients will receive
absorbed doses to malignant and normal tissues that are excessively higher than
necessary. Undertreatment will result in further administrations of radioiodine
with the risk of dedifferentiation over time, so that tumours become less iodine
avid. Overtreatment can result in unnecessary toxicity which can take the
form of sialadenitis and pancytopenia. Radiation pneumonitis and pulmonary
fibrosis have been seen in patients with diffuse lung metastases, and there is a
risk of leukaemia in patients receiving high cumulative activities. Personalized
treatments were first explored in the 1960s with patients administered activities
required to deliver a 2 Gy absorbed dose to the blood and constraints regarding
radioactive uptake levels at 48 h. Further approaches have been taken, based on
whole body absorbed doses, which can be considered a surrogate for absorbed
doses to the red marrow.
Dosimetry for thyroid ablations presents a different set of challenges to
that performed for therapies. In the former case, the small volume of remnant
tissue can render delineation inaccurate, which subsequently impinges on the
accuracy of the dose calculation. Therapies of metastatic disease can involve
larger volumes, although heterogeneous uptake is frequently encountered and
lung metastases in particular require careful image registration and attenuation
correction (Fig.19.1).
A current issue is that of stunning, whereby a tracer level of activity may
mitigate further uptake for an ablation or therapy. This phenomenon would have
consequences for individualized treatment planning, although at present its extent
and indeed its existence is being contested. However, it is not infrequent that a
greater extent of uptake may be seen from a larger therapy administration than
from a tracer administration (Fig.19.1).
While subject to national regulations, patients receiving radioiodine
treatment frequently require in-patient monitoring until retention of activity falls
to levels acceptable to allow contact with family members and the public. It is,
644
RADIOnUCLIDE tHERAPY
therefore, necessary for the physicist to give strict advice on radiation protection,
taking into account the patients home circumstances.
FIG. 19.1. Absorbed dose maps resulting from a tracer administration of 118 MBq 131I-NaI
(left) and, subsequently, 8193 MBq 131I-NaI (right) for therapy (maximum absorbed dose:
90 Gy). The absorbed doses were calculated using 3-D dosimetry on a voxel by voxel basis.
19.3. PalliaTioN of boNe PaiN

bony metastases arise predominantly from prostate and breast cancer. bone
pain is experienced by up to 90% of patients with castration resistant prostate
cancer in the later phases of their disease. radiopharmaceuticals have been
established as an effective agent for bone pain palliation for almost 70 years,
with 89sr first being used in 1942. a wide range of radiopharmaceuticals have
been used to treat bone metastases and there are two commercially available
products, 89sr chloride (Metastron) and 153sm-lexidronam (Quadramet) that have
received us food and drug administration (fda) approval (in 1993 and 1997,
respectively). a number of other radiopharmaceuticals have been used, including
32
P, 186re-hedP, 188re-hedP, 117msn and 177lu-edTMP. More recently, the
emitter 223ra has undergone evaluation in randomized phase iii clinical trials
and has also received fda approval.
in the case of 89sr and 153sm, administered activities tend to be standardized
according to the manufacturers guidelines. however, administered activities
for other agents vary widely according to local protocols, and published
guidelines are largely concerned with procedure. re-treatments are generally
considered to be beneficial, subject to recovery of haematological toxicity, and
recommendations for the timing of these have been made by both the eaNM
645
CHAPTER 19
and the IAEA. However, no trials have yet been performed to assess the optimal
timing or levels of administration.
19.3.1. Treatment specific issues
The main issue concerning the use of radiopharmaceuticals for the treatment
of bone pain is that of determining the ideal treatment protocol, including the
optimal radionuclide to use, and whether this should be standardized or could be
modified on an individual patient basis. In practice, local logistics and availability
will have a strong impact on the radionuclide of choice. It is of particular interest
that the radionuclides used vary widely in terms of their emissions. Arguments
can be made to support both approaches, in that the longer range emitters may
be more likely to target all of the disease, while the shorter range emitters
(and particularly an emitter) will avoid unnecessary toxicity. There is also a
wide range of physical half-lives between these radionuclides and there is some
evidence to suggest that the longer lived 89Sr can produce a response that takes
longer to occur but that is longer lasting.
Dosimetry for bone pain palliation is challenging due to the difficulties
of assessing the distribution of uptake in newly formed trabecular bone and its
geometrical relation to viable red marrow and to disease. Nevertheless, models
have been developed to address this interesting problem and a statistically
significant correlation has been demonstrated between whole body absorbed
doses and haematological toxicity. Dosimetry for other radionuclides is highly
dependent on the imaging properties of these radionuclides, although it could
potentially be used to increase administered activities in individual patients.
19.4. HEPATIC CANCER
Hepatocellular carcinoma is a major cause of cancer deaths. In recent
years, primary and secondary liver cancers have been treated with a range of
radionuclides administered intra-arterially, based on the fact that while the
liver has a joint blood supply, tumours are supplied only by the hepatic artery.
The advantage of this approach is that treatments can be highly selective and
canminimize absorbed doses delivered to normal organs, including healthy
liver. This procedure requires interventional radiology as the activity must
be administered directly into the common, right or left hepatic artery via an
angiographic catheter under radiological control and so is a prime example of
the multidisciplinary nature of radionuclide therapy. Prior to administration,
a diagnostic level of 99mTc-macroaggregate of albumin (MAA) is given to
646
ascertain the likelihood of activity shunting to the lung. This is usually evaluated
semi-quantitatively.
To date, two commercial products have received FDA approval, classified
as medical devices rather than as drugs. Both use 90Y. Theraspheres comprise 90Y
incorporated into small silica beads and SIR-Spheres consist of 90Y incorporated
into resin. Lipiodol, a mixture of iodized ethyl esters of the fatty acids of poppy
seed oil, has also been explored for intra-arterial administration. Lipiodol
has been radiolabelled with both 131I and 188Re, the latter having the benefit
of superior imaging properties, a longer path length and fewer concerns for
radiation protection due to the shorter half-life.
As with other therapies, outstanding issues include the optimal activity
to administer, which is usually based on patient weight or body surface area,
arteriovenous shunting observed prior to treatment and the extent of tumour
involvement. However, there have been examples of treatments planned
according to estimated absorbed doses delivered to the normal liver and this
treatment offers the potential for individualized treatment planning based on
potential toxicity. Radiobiological consequences have been considered by
conversion of absorbed doses to biologically effective doses and there are
tentative conclusions that multiple treatments may deliver higher absorbed doses
to tumours whileminimizing absorbed doses to normal liver.
A particular issue of this treatment concerning the physicist is that
of imaging, due to the need to ascertain lung uptake from the pre-therapy
99m
Tc-MAA scan, and the possibility of bremsstrahlung imaging as a basis for
calculation of absorbed doses delivered at therapy.
19.5. NEUROENDOCRINE TUMOURS
Neuroendocrine tumours (NETs) arise from cells that are of neural crest
origin and usually produce hormones. There are several types of neuroendocrine
cancer, including phaeochromocytoma, which originates in the chromaffin cells
of the adrenal medulla, and paraganglioma, which develops in extra-adrenal
ganglia, often the abdomen. Carcinoid tumours are slow growing and arise
mainly in the appendix or small intestine although they can also be found in
the lung, kidney and pancreas. Medullary thyroid cancer is a special case of an
NET that arises from the parafollicular cells of the thyroid gland, which produce
calcitonin. For the purposes of radionuclide therapy, NETs tend to be considered
as one malignancy and similar radiopharmaceutical treatments are administered,
647
CHAPTER 19
although due in part to differences in radiosensitivity and proliferation,

response is variable between diseases. NETs are frequently treated with
radiopharmaceuticals. The mean age at diagnosis is around 60 years although
tumours may present at any age.
There are two main mechanisms by which NETs are targeted with
radiopharmaceuticals. NETs have been treated with the noradrenaline analogue
131
I-MIBG (metaiodobenzylguanidine) for over twenty years, although this is
still largely considered an experimental treatment. Although generally considered
to be a palliative treatment, high uptake can be achieved and complete responses
have been seen. More recently, a number of peptide analogues of somatostatin
have been developed, radiolabelled with 90Y, 111In or 177Lu, that offer a range
of treatment options. Guidelines for radionuclide therapy of NETs have been
produced by the EANM and the European Neuroendocrine Tumour Society,
and focus mainly on procedural aspects. Recommendations are not given
for administered activities, and these can vary from 3700 to 30000MBq of
131
I-MIBG and cumulated activities of 1200018000MBq of 90Y-DOTATOC.
Administrations are often repeated, although there are no standardized protocols
for the intervals between therapies.
To date, there have been almost no studies directly comparing the relative
merits of the different radiopharmaceuticals available for the treatment of
neuroendocrine cancer. Key considerations are largely related to the relative
path lengths of the radionuclides used, their imaging properties and toxicity.
For example, 111In-octreotide therapy readily lends itself to imaging due to dual
emission peaks at 173 and 247keV, and relies on internalization due to radiation
delivered by Auger emissions, whereas 90Y labelled analogues can cause
irradiation over 1cm although they can only be imaged with bremsstrahlung
scintigraphy. 90Y labelled analogues and 131I-MIBG can cause myelosuppression,
thus the administration of higher activities may require stem cell support. Kidney
toxicity can be another activity-limiting factor for the somatostatin analogues.
The range of activities administered and the increasingly available range
of radiopharmaceuticals developed for the treatment of NETs is indicative
of the main issue facing this treatment, which is to determine the optimal
treatment protocol. As with other therapies, the issue of personalized versus
standardized treatments is at the forefront of this debate, with administrations
based on fixed activities modified according to patient weight or, in some cases,
based on absorbed whole body doses. It has been shown that a wide range of
absorbed doses are delivered to both tumours and to normal organs from fixed
activities. Dosimetry for high activities of 131I-MIBG has been the subject of
648
extensive research in recent years, and must deal with problems resulting from
camera dead time, photon scatter and attenuation. Image based dosimetry of
90
Y labelled pharmaceuticals has been performed using low levels of 111In given
either prior to therapy or included with the therapy administration. More recently,
bremsstrahlung imaging has been developed to enable dosimetry to be performed
directly.
19.6. NON-HODGKINS LYMPHOMA
Of the malignancies arising from haematological tissues, non-Hodgkins
lymphoma is most commonly targeted with radiopharmaceuticals. Various forms
of lymphoma are classified into high grade or low grade, depending on the rate of
growth. Lymphomas are inherently radiosensitive, express a number of antigens
and can be successfully treated with radioimmunotherapy (RIT) using monoclonal
antibodies radiolabelled usually with either 131I or 90Y. A number of radiolabelled
monoclonal antibodies have been developed and two, 90Y-Ibritumomab Tiuxitan
(Zevalin) and 131I-Tositumomab (Bexxar), have received FDA approval. Both
target the B-cell specific CD 20 antigen and have been used successfully in a
number of clinical trials. Both agents have demonstrated superior therapeutic
efficacy to prior chemotherapies in various clinical settings.
As with chemotherapy, RIT is more successful when administered at an
early stage of disease. Clinical trials are ongoing to determine how to more
effectively integrate RIT into the current clinical management algorithm in
lymphoma patients.
Internal dosimetry has been applied to a number of studies using RIT with
varying results and conclusions. Initial dosimetry trials for Zevalin found that
while absorbed doses were delivered to tumours and to critical organs that varied
by at least tenfold, these did not correlate with toxicity or response, and that at
the levels of activity prescribed, the treatment was deemed to be safe, obviating
the need for individualized dose calculations, and treatment now tends to be
administered based on patient weight. However, FDA approval incorporated
the need for biodistribution studies to be performed prior to therapy using the
antibody radiolabelled with 111In as a surrogate for 90Y under the assumption that
the tracer kinetics would translate into clinical therapy and a number of studies
are now concerned with assessing the biodistribution and dosimetry based on
bremsstrahlung imaging (Fig.19.2).
649
CHAPtER 19
FIG. 19.2. An absorbed dose map (maximum dose: 39 Gy) resulting from 3-D dosimetry
of bremsstrahlung data acquired from treatment of non-Hodgkins lymphoma with
90
Y-Ibritumomab Tiuxitan (Zevalin).
in contrast, studies using 131i-Tositumomab (bexxar) have demonstrated

that at least bone marrow toxicity is significantly related to dosimetry. as a
result, 131i-Tositumomab is one of the few radionuclide therapies (or indeed
radiotherapy procedures) based on individualizing absorbed doses delivered to
the critical organ, which in this case is the bone marrow. To this end, the level of
administered activity is determined according to a whole body absorbed dose of
0.75 Gy, calculated from a series of three whole body scintigraphy scans.
19.7. PaediaTric MaliGNaNcies
cancer in children is rare, with an incidence of less than 130 per million
and an overall relative survival rate of 57%. leukaemia and lymphoma account
for nearly 50% of cases. radionuclide therapy for children and young people is
correspondingly rare and entails particular scientific and logistical challenges that
justify consideration independently of adult treatments. issues of in-patient care
predominantly arise due to the combination of increased nursing requirements
and radiation protection considerations. radiation protection must also play a
large role in decisions to allow children to leave hospital, as they frequently have
siblings at home.
650
19.7.1. Thyroid cancer

The ablation and therapy of thyroid cancer with radioiodine is performed for
children, who are considered a high risk group. There is commonly a significantly
higher incidence of metastatic disease in children than in adults. Fatalities can
be as high as 25%, often occurring after many years of repeated radioiodine
treatments and consequently high cumulated activities. Thus, potential late
toxicity in children from radionuclide therapy needs to be considered.
19.7.2. Neuroblastoma
Neuroblastoma is a malignancy of the neuroendocrine system specific to
children and young people. Neuroblastoma is inherently radiosensitive and has
been treated with 131I-MIBG since the 1980s, particularly for primary refractory
or relapsed patients. Although treatments are generally intended to be palliative,
complete responses have been reported. More recently, there has been interest in
treating neuroblastoma with radiolabelled peptides, such as 177Lu-Dotatate.
While it is generally recognized that protocols for administering radioiodine
to children should be modified from those applied to adults, there is little
agreement on how such modifications should be determined and, in practice,
these can be based on body weight, surface area or age. The EANM guidelines
on radioiodine therapy of differentiated thyroid cancer support the principle
of individual treatment of thyroid cancer in children and German procedure
guidelines advocate administration based partly on the 24h uptake of a tracer
activity prior to ablation.
Despite the small number of centres that have treated children with
131
I-MIBG for the treatment of neuroblastoma, there has been a very wide
variation in treatment protocols. While many treatments have relied on fixed
activities (which have generally ranged from 3700 to 7400MBq), substantial
research and development into quantitative imaging and internal dosimetry of
131
I has led to a higher degree of dosimetry based personalized treatments than
has been the case for adult therapies. This has led in particular to administered
activities being calculated based on whole body absorbed doses which have
been shown to correlate with haematological toxicity. Further complications,
particularly relating to toxicity and radiation protection, can be caused by an
increasing trend towards higher activities, possibly administered with stem cell
support and concomitant chemotherapy which can act as a radiosensitizer.
651
CHAPTER 19
19.8. ROLE OF THE PHYSICiST

The physicist is responsible for a wide range of tasks in nuclear medicine
and must perform duties that include the procurement and maintenance of
imaging equipment and associated computer systems; responsibility for radiation
protection and interpretation; and implementation of national legislation. This
comprehensive and challenging role is exemplified in the treatment of cancer and
benign disease with radiopharmaceuticals that can entail levels of radioactivity
far exceeding those used for diagnostic purposes.
Quality control of scintillation cameras is fundamental to good clinical
practice in radionuclide therapy to ensure that the diagnostic information used as
a basis for treatment is accurate. This relies on the development of and adherence
to a strict procedure of well defined protocols and procedures that must be
performed regularly.
Radiation protection pertaining to unsealed sources used for radionuclide
therapy entails a greater degree of responsibility than that commonly encountered
in diagnostic imaging. Staff are potentially exposed to high levels of radiation
of , and emissions which, therefore, requires careful handling, dispensing
and administration of therapeutic radiopharmaceuticals. Therapy procedures can
involve staff groups that usually do not encounter high levels of radiation, so
that extra precautions are needed. This particularly applies to the care of patients,
which may be provided to some extent by family members and carers as well
as by nurses, and the scanning of patients following administrations of high
activities by radiographers and technicians. Careful monitoring of the involved
staff must be performed at all times and the physicist must be aware of national
regulations.
There is an increasing opportunity for development of a number of related
areas in radionuclide therapy which predominantly involve the physicist.
Foremost among these is quantitative imaging. While the clinical viewpoint of
radionuclide therapy is focused on the indication and on the treatment, imaging
concerns only matters that, to some extent, may be independent of these. Thus,
for example, quantitative imaging of 131I uptake in the abdomen will follow
the same procedure whether this results from 131I-MIBG treatment of an NET
or an 131I-radiolabelled monoclonal antibody for the treatment of lymphoma.
Optimization of imaging of bony metastases with a given agent will follow
similar procedures whether the metastases arise from prostate or breast cancer.
Quantitative imaging is, therefore, predominantly focused on the radionuclide,
independently of its formulation, and on the extent and localization of uptake as
well as the imaging equipment employed for the purpose. Quantitative imaging
must take into account a number of factors that are often of little concern in
diagnostic imaging. Scatter is a significant impediment to accurate quantitative
652
imaging, particularly where high energy emitters such as 131I are used for therapy.
Corrections can be applied with relative ease by assessing and subtracting the
scatter contribution from one or more energy windows placed adjacent to the
peak energy window. Attenuation correction is essential to quantitative imaging
and can be performed using a variety of methods. These can range from a
straightforward approach that assumes the patient consists entirely of water,
to more sophisticated methods that take into account the electron density on a
voxel by voxel basis. Dead time corrections are frequently overlooked in the
imaging of patients undergoing radionuclide therapy as these seldom require
consideration for diagnostic scanning. However, this is an issue of paramount
importance that will severely inhibit accurate quantification if ignored. Again,
this is a particularly significant factor when using high activities of 131I, and it is
essential that each camera is characterized accordingly prior to image processing
and analysis.
Accurate image quantification enables further avenues of research and
development that have only recently begun to emerge as substantial areas of study.
Pharmacokinetic analysis, derived from sequential scanning, can allow inter- and
intra-patient variations in uptake and retention to be calculated which can aid
understanding and optimization of a radiopharmaceutical. This is particularly
relevant for new products. Accurate analysis is dependent on the acquisition of
sufficient statistics and data which must include the number and timing of scans.
Inherent errors and uncertainties should be considered where possible.
Quantitative imaging and analysis facilitate the accomplishment of
accurate internal dosimetry calculations, which are of paramount importance to
patient specific treatment planning and which are now becoming mandatory for
new products and necessary for the acquisition of the evidence base on which
treatments should be performed. Although isolated studies into image based,
patient specific dosimetry have been performed for many years, this remains a
newly emerging field for which no standardized protocols or guidelines exist.
This puts greater responsibility on the physicist whose role must be to advise
on image acquisition. This invariably entails balancing scientific requirements
with local resource restrictions. As there is only limited software available for
dosimetry calculations at present, it may prove necessary to develop software or
spreadsheets to perform absorbed dose calculations.
Essentially, the end point of such calculations should lead to a prediction of
absorbed doses to the tumour and critical organs from a given administration and
confirmation of the absorbed doses delivered after the therapy has been performed.
However, interpretation and understanding of the biological relevance of these
absorbed doses is not straightforward. Radiobiology for radionuclide therapy has
not been developed at the rate seen for external beam radiotherapy (EBRT) but
is now attracting more attention. There are both biological and physics aspects
653
CHAPTER 19
to radiobiology, with the latter largely concerned with constructing models to

explain physiological phenomena. To some extent, these models may be adapted
from those formulated for EBRT, which are predominantly based on the so-called
linear quadratic model. This model is largely predicated on the assumption that
cellular radiation damage can be considered separately according to single or
double strand DNA (deoxyribonuleic acid) breaks and while this has a validity
for radionuclide therapy, a number of confounding factors should be taken into
account that can accommodate the relatively low but continuous absorbed dose
rates delivered by radioactive uptake and emerging evidence to suggest that DNA
is not the only target causing cell death. Radiobiology for radionuclide therapy
is likely to become more complicated as radiopharmaceuticals are administered
with concomitant chemotherapy or with EBRT and as new factors are discovered,
such as the bystander effect, in which unirradiated cells can be killed if they are
in proximity to cells that have been irradiated, and hyper-radiosensitivity, which
has indicated excessive sensitivity to very low levels of irradiation.
In summary, it is likely that the varied role enjoyed by the physicist will
become more complicated as radionuclide therapies are increasingly subject to
accountability and as the field expands. Radionuclide therapy is the only cancer
treatment modality that allows imaging of the therapeutic drug in situ. It is the
duty of the physicist to capitalize on this by providing the information necessary
to enable optimal and cost effective treatment.
19.9. EMERGING TECHNOLOGY
In the coming years, a number of factors will affect the development of
radionuclide therapy and, in particular, the role of the physicist. New imaging
technology has emerged in recent years in the form of hybrid scanners, which
will have a significant impact on improving the accuracy of dosimetry from
radionuclide therapies.
A range of new radiopharmaceuticals is now emerging and previously
used radiopharmaceuticals are being revisited. In particular, there is currently a
growing interest in emitters, although these have been used in some form for
many decades. The physical properties of particles have distinct advantageous
and disadvantageous implications for radionuclide therapy. The short range
of emissions (10100 m in soft tissue) means that uniform uptake of a
radiopharmaceutical is critical to a successful treatment as there is little radiation
crossfire. However, the high linear energy transfer ensures that radiation damage
resulting from uptake in a cell is likely to be lethal and that cells immediately
adjacent are also likely to be killed. Examples of alpha therapy to date include
211
At for direct infusion into resected gliomas, antibodies radiolabelled with
654
213
Bi or 225Ac for the treatment of leukaemia and 223Ra for the treatment of bone
metastases. Dosimetry for emitters remains largely unexplored and is subject to
a number of challenges due to the difficulty of localization and the need to take
into account the emissions of daughter products, which may not remain at the
initial site of uptake.
The introduction of more stringent regulatory procedures will increase
the need for accurate internal dosimetry. The FDA now requires dosimetric
evaluation of new radiopharmaceuticals, and it is becoming commonplace for
new uses of existing agents to also be the subject of a phase I/II clinical trial to
ascertain absorbed doses delivered to critical organs. As brief palliative effects
translate into longer lasting survival, critical organ dosimetry will become more
important to ensureminimization of unnecessary late toxicity. However, it should
be noted that basic dosimetry calculations aimed at estimating critical organ
absorbed doses are not necessarily sufficient to ensure an optimal treatment
protocol, which must take tumour dosimetry into account.
The increasing trend towards accountability and evidence based medicine
will require adherence to strict radiation protection procedures for patients,
families and staff, and it may become necessary to assess exposure, particularly
to family members, with greater accuracy.
Scientific developments are likely to proceed rapidly, and many are now
within the reach of departments that have only basic research facilities, since
accurate absorbed dose calculations can be obtained from careful imaging
procedures and a relatively simple spreadsheet. Individualization of absorbed
dose calculations can then be achieved. Dosimetry based treatment planning
will also become an essential element of patient management as options of
chemotherapy or EBRT administered concomitantly with radiopharmaceuticals
are explored. The practice of internal dosimetry itself continues to evolve and
can be divided into categories that require different approaches. In addition
to image based dosimetry, these include whole body dosimetry, blood based
dosimetry and model based dosimetry. A particular focus at present is on
red marrow dosimetry, as this is the absorbed dose limiting organ for many
therapies.
Multi-centre prospective data collection is critical to the development
of this field, and international networks will be required to accrue a sufficient
number of patient statistics to enable the formulation of agreed and standardized
treatment protocols.
655
CHAPTER 19
19.10. CONCLUSIONS
Nuclear medicine physics is playing an increasingly important role
in the service and management of radionuclide therapies. In addition to the
tasks traditionally associated with nuclear medicine, which primarily involve
the maintenance of imaging and associated equipment, radiation protection
and administration of national regulations, there is a growing requirement for
patient specific treatment planning which requires quantitative imaging, internal
dosimetry and radiobiological considerations.
While radionuclide therapy is usually performed within nuclear medicine, it
is often to be found within clinical or medical oncology, or within endocrinology.
In effect, radionuclide therapy requires a multidisciplinary approach that involves
diverse groups of staff. The adoption of treatments based on individualized
biokinetics, obtained from imaging and external retention measurements, places
the physicist more centrally within this network, as is seen in EBRT.
There is currently the need for increased training in this field, and due
to the relatively low numbers of patients treated even at specialist centres,
multi-centre networks will facilitate the exchange of expertise and the gathering
of prospective data necessary to advance the field. As this hitherto overlooked
area of cancer management expands, the scientific opportunities available to the
nuclear medicine physicist will also increase.
BIBLIOGRAPHY
BUFFA, F.M., et al., A model-based method for the prediction of whole-body absorbed dose
and bone marrow toxicity for Re-186-HEDP treatment of skeletal metastases from prostate
cancer, Eur. J. Nucl. Med. Mol. Imaging 30 (2003) 11141124.
CREMONESI, M., et al., Radioembolisation with Y-90-microspheres: dosimetric and
radiobiological investigation for multi-cycle treatment, Eur. J. Nucl. Med. Mol. Imaging 35
(2008) 20882096.
DU, Y., HONEYCHURCH, J., JOHNSON, P., GLENNIW, M., ILLIDGE, T., Microdosimetry
and intratumoral localization of administered 131I labelled monoclonal antibodies are critical to
successful radioimmunotherapy of lymphoma, Cancer Res. 67 (2003) 13351343.
GAZE, M.N., et al., Feasibility of dosimetry-based high-dose I-131-meta-iodobenzylguanidine
with topotecan as a radiosensitizer in children with metastatic neuroblastoma, Cancer Biother.
Radiopharm. 20 (2005) 195199.
LASSMANN, M., LUSTER, M., HANSCHEID, H., REINERS, C., Impact of I-131 diagnostic
activities on the biokinetics of thyroid remnants, J. Nucl. Med. 45 (2004) 619625.
656
MADSEN, M., PONTO, J., Handbook of Nuclear Medicine, Medical Physics Publishing,
Madison, WI (1992).
NINKOVIC, M.M., RAICEVIC, J.J., ADROVIC, F., Air kerma rate constants for gamma
emitters used most often in practice, Radiat. Prot. Dosimetry 115 (2005) 247250.
STOKKEL, M.P., HANDKIEWICZ JUNAK, D., LASSMANN, M., DIETLEIN, M.,
LUSTER, M., EANM procedure guidelines for therapy of benign thyroid disease, Eur. J. Nucl.
Med. Mol. Imaging 37 (2010) 22182228.
657
CHAPTER 20
MANAGEMENT OF THERAPY PATIENTS
L.T. DAUER
New York, United States of America
20.1. INTRODUCTION
The basic principles of radiation protection and their implementation as
they apply to nuclear medicine are covered in general in Chapter3. This chapter
will look at the specific case of nuclear medicine used for therapy. In addition
to the standards discussed in Chapter3, specific guidance on the release of
patients after radionuclide therapy can be found in the IAEAs Safety Reports
Series No. 63 [20.1].
When the patient is kept in hospital following radionuclide therapy, the
people at risk of exposure include hospital staff whose duties may or may not
directly involve the use of radiation. This can be a significant problem. However,
it is generally felt that it can be effectively managed with well trained staff and
appropriate facilities. On the other hand, once the patient has been released, the
groups at risk include members of the patients family, including children, and
carers; they may also include neighbours, visitors to the household, co-workers,
those encountered in public places, on public transport or at public events, and
finally, the general public. It is generally felt that these risks can be effectively
mitigated by the radiation protection officer (RPO) with patient-specific radiation
safety precaution instructions.
20.2. OCCUPATIONAL EXPOSURE
20.2.1. Protective equipment and tools
Protective clothing should be used in radionuclide therapy areas where
there is a likelihood of contamination. The clothing serves both to protect the
body of the wearer and to help to prevent the transfer of contamination to other
areas. Protective clothing should be removed prior to going to other areas such as
staff rooms. The protective clothing may include laboratory gowns, waterproof
658
gloves, overshoes, and caps and masks for aseptic work. When emitters are
handled, the gloves should be thick enough to protect against external radiation
(perhaps double gloves should be utilized, when appropriate).
In radionuclide therapy nuclear medicine, most of the occupational
exposures come from 131I, which emits 356keV photons. The attenuation by a
lead apron at this energy isminimal (less than a factor of two) and is unlikely to
result in significant dose reductions and may not justify the additional weight and
discomfort of wearing such protective equipment. Typically, thicker permanent
or mobile lead shielding may be more effectively applied for those situations
which warrant its use. The RPO should determine the need and types of shielding
required for each situation.
20.2.2. Individual monitoring
Individual monitoring, as discussed in Chapter3, needs to be considered
during the management of radionuclide therapy patients. In addition to general
advice (seeChapter3) on persons most likely to require individual monitoring in
nuclear medicine, consideration needs to be given to nursing or other staff who
spend time with therapy patients.
20.3. RELEASE OF THE PATIENT
Protection of the patient in therapeutic nuclear medicine is afforded
through the application of the principles of justification and optimization the
principle of dose limitation is not applied to patient exposures. A discussion of
these principles is given in Chapter3. However, a patient that has undergone a
therapeutic nuclear medicine procedure is a source of radiation that can lead to
the exposure of other persons that come into the proximity of the patient. External
irradiation of the persons close to the patient is related to the radionuclide used,
its emissions, half-life and biokinetics, which can be important with some
radionuclides. Excretion results in the possibility of contamination of the patients
environment and of inadvertent ingestion by other persons.
The system of radiation protection handles, in different ways, people that
may be exposed by therapeutic nuclear medicine patients. If the person is in close
proximity because their occupation requires it, then they are subject to the system
of radiation protection for occupationally exposed persons. If the person, other
than occupationally, is knowingly and voluntarily providing care, comfort and
support to the patient, then their exposure is considered part of medical exposure,
and they are subject to dose constraints as discussed in Chapter3. If the person
is simply a member of the public, including persons whose work in the nuclear
659
CHAPTER 20
medicine facility does not involve working with radiation, then their exposure is
part of public exposure and that is discussed in the next section.
While precautions for the public are rarely required after diagnostic nuclear
medicine procedures, some therapeutic nuclear medicine procedures, particularly
those involving 131I, can result in significant exposure to other people, especially
those involved in the care and support of patients. Hence, members of the public
caring for such patients in hospital or at home require individual consideration.
20.3.1. The decision to release the patient
Patients do not need to be hospitalized automatically after all radionuclide
therapies. Relevant national dose limits must be met and the principle of
optimization of protection must be applied, including the use of relevant dose
constraints. The decision to hospitalize or release a patient should be determined
on an individual basis. In addition to residual activity in the patient, the decision
should take many other factors into account. Hospitalization will reduce
exposure to the public and relatives, but will increase exposure to hospital staff.
Hospitalization often involves a significant psychological burden as well as
monetary and other costs that should be analysed and justified.
Medical practitioners shall determine whether the patient is willing and
is physically and mentally able to comply with appropriate radiation safety
precautions in the medical facility, should medical confinement be necessary, or
at home after release. For some patients, hospitalization during and following
treatment may be necessary and appropriate. The medical practitioners can
determine that such patients may need to remain hospitalized beyond the period of
time dictated by other dose constraint or clinical criteria. For example, incontinent
patients or ostomy patients may require extended hospitalization to ensure safe
collection and disposal of radioactively contaminated body wastes. Where the
social system and infrastructure is such that there may be contamination risks
from discharged patients, it may be necessary to hospitalize the patient or extend
the normal hospitalization time, to avoid risk to the environment or other persons
[20.1].
The decision to hospitalize or release a patient after therapy should be made
on an individual basis considering several factors including residual activity in
the patient, the patients wishes, family considerations (particularly the presence
of children), environmental factors, and existing guidance and regulations. The
nuclear medicine physician has the responsibility to ensure that no patient who
has undergone a therapeutic procedure with unsealed sources is discharged from
the nuclear medicine facility until it has been established by either a medical
physicist or by the facilitys RPO that the activity of radioactive substances in
the body is such that the doses that may be received by members of the public
660
and family members would meet national criteria, including compliance with
relevant dose limits and the application of relevant dose constraints. Iodine-131
typically results in the largest dose to medical staff, the public, caregivers and
relatives. Other radionuclides used in therapy are usually simple emitters
(e.g. 32P, 89Sr and 90Y) that pose much less risk.
The modes of exposure to other people are: external exposure, internal
exposure due to contamination, and environmental pathways. The dose to
adults from patients is mainly due to external exposure. Internal contamination
of family members is most likely in the first seven days after treatment. In
most circumstances, the risks from internal contamination of others are less
significant than those from external exposure [20.1]. In general, contamination
of adults is less important than external exposure. However, contamination of
infants and children with saliva from a patient could result in significant doses
to the childs thyroid [20.2]. Therefore, it is important to avoid contamination
(particularly from saliva) of infants, young children and pregnant women owing
to the sensitivity of fetal and paediatric thyroids to cancer induction [20.1,
20.3]. Written instructions to the patient concerning contact with other persons
and relevant precautions for radiation protection must be provided as necessary
(seeSection 20.3.2).
The day to day management of hospitalization and release of patients should
be the responsibility of the licensee. In applying dose constraints, registrants and
licensees should have a system to measure or estimate the activity in patients
prior to discharge and assess the dose likely to be received by members of the
household and members of the public. The result should be recorded. A method to
estimate the acceptable activity of radiopharmaceuticals for patients on discharge
from hospitals is to calculate the time integral of the ambient dose equivalent rate
and compare it with the constraints for patient comforters, or for other persons
who may spend time close to the patient. For this calculation, either a simple
conservative approach based on the physical half-life of the radionuclide or a
more realistic one, based on patient-specific effective half-life, can be used. The
assumptions made in these calculations with regard to time (occupancy factors)
and distance should be consistent with the instructions given to patients and
comforters at the time the patient is discharged from hospital. In the calculation
of the effective half-life, the behaviour of 131I can be modelled using two
components for the biological half-life: the extra-thyroidal (i.e. existing outside
the thyroid) iodine and thyroidal iodine following uptake by thyroid tissue. The
assumptions used often err on the side of caution; it is sometimes felt that they
significantly overestimate the potential doses to carers and the public. Examples
of such calculations are found in the literature [20.4, 20.5]. Further guidance on
radiation protection following radionuclide therapy can be found in Ref.[20.1]
(especially in annex II).
661
CHAPTER 20
When deciding on the appropriate discharge activity for a particular patient,

the licensee should take into account the transport and the living conditions of the
patient, such as the extent to which the patient can be isolated from other family
members and the requirement to dispose safely of the patients contaminated
excreta. Special consideration shall be given to the case of incontinent patients.
In some cases, such as for the elderly or children, it may be necessary to discuss
the precautions to be taken with other family members.
Additional guidance on specific release considerations depending on
various radionuclide therapies can be found in annexV of Ref.[20.1].
20.3.2. Specific instructions for releasing the radioactive patient
Current recommendations regarding release of patients after therapy with
unsealed radionuclides vary widely around the world. However, the decision to
release a patient is based on the assumption that the risk can be controlled when
the patient returns to their home. This is generally achieved by combining an
appropriate release criterion with well tailored instructions and information for
the patient that will allow them to deal effectively with the potential risk [20.1].
When required, the patient or legal guardian shall be provided with
written (and perhaps a verbal explanation of) instructions with a view to the
restriction of doses to persons in contact with the patient as far as reasonably
achievable, and information on the risks of ionizing radiation. It is important
to develop effective communication methods. The IAEA gives example
information/leaflet information in Safety Reports Series No. 63 [20.1]. Specific
instructions should include items such as instructions to patients concerning
the spread of contamination,minimization of exposure to family members,
cessation of breast-feeding, and conception after therapy. The amount of time
that each precaution should be implemented should be determined based on an
estimate of the activity in patients prior to discharge and an assessment of the
dose likely to be received by carers and comforters or members of the public
under various precaution formulations as compared to the appropriate dose
constraints. Procedures for advising carers and comforters should be in place in
consultation with the RPO. Registrants and licensees should ensure that carers
and comforters of patients during the course of treatment with radionuclides (e.g.
131
I for hyperthyroidism and thyroid carcinoma; 89Sr, 186Re for pain palliation)
receive sufficient written instructions on relevant radiation protection precautions
(e.g. time and proximity to the patient). Example methodologies for evaluating
precaution time requirements have been published [20.5, 20.6].
Female patients should be advised that breast-feeding is contraindicated
after therapeutic administration of radionuclides, and females as well as males
should be advised concerning the avoidance of conception after therapeutic
662
administrations. The IAEAs Safety Reports Series No. 40 [20.7] recommends

cessation of breast-feeding for a patient given 5550MBq (150 mCi) of
131
I-NaI. Following treatment with a therapeutic activity of a radionuclide, female
patients should also be advised to avoid pregnancy for an appropriate period.
The International Commission on Radiological Protection (ICRP) suggests that
women should not become pregnant for some time after radionuclide therapy
(e.g. 6 months for radioiodine, the most common radionuclide used) [20.2].
Various shorter or longer times for this and other radionuclides are given in
ICRP Publication 94 [20.8] and Ref.[20.7] which identifies periods of 3, 4 and
24 months for 32P, 131I and 89Sr treatments, respectively. Some practitioners use
a 612 month gap for 131I, with a view to providing further confidence in this
regard. Table13 of Ref.[20.1] gives additional information on precaution times
for female avoidance of conception for specific radionuclide therapies [20.1].
The administration of therapeutic doses of relatively long lived radionuclides
in ionic chemical forms to males is also a possible source of concern because
of the appearance of larger quantities of these radionuclides in ejaculate and in
sperm. It is widely recommended in practice, on the basis of prudence, that male
patients take steps to avoid fathering children during the months immediately
following therapy [20.1]. However, there is no strong evidence base to support
this view. Some have suggested that it may be prudent to advise sexually active
males who have been treated with 131I (iodide), 32P (phosphate) or 89Sr (strontium
chloride) to avoid fathering children for a period of 4 months after treatment, a
period suggested as it is longer than the life of a sperm cell [20.9].
Patients travelling after radioiodine therapy rarely present a hazard to other
passengers if travel times are limited to a few hours. Travel for 12h immediately
post-treatment in a private automobile large enough for the patient to maintain
a distance of 1 m or greater from the other vehicle occupant(s) is generally
permissible. A case by case analysis is necessary to determine the actual travel
restrictions for each patient, especially for longer trips and for travel by public
transport.
Current international security measures, such as those in place at airports
and border crossing points, can include extremely sensitive radiation detectors. It
is quite possible that patients treated with emitting radionuclides could trigger
these alarms, particularly in the period immediately following discharge.
Environmental or other radiation detection devices are able to detect patients
who have had radioiodine therapy and some diagnostic procedures for several
weeks after treatment [20.10, 20.11]. Triggering of an alarm does not mean that a
patient is emitting dangerous levels of radiation the detectors are designed to
detect levels of radioactivity far below those of concern to human health. The
security authorities are well aware of this possibility, and if a patient is likely to
travel soon after discharge, the hospital or the patients doctor should provide a
663
CHAPTER 20
written statement of the therapy and radionuclide used, for the patient to carry.
The IAEA gives an example of a credit card-style card that might be given to a
patient at the time of discharge (seeFig.20.1) [20.1]. Personnel operating such
detectors should be specifically trained to identify and deal with nuclear medicine
patients. Records of the specifics of therapy with unsealed radionuclides should
be maintained at the hospital and given to the patient along with written
precautionary instructions [20.2].
(a)
(b)
FIG.20.1. Example of a credit card-style card that might be given to a patient at the time of
discharge: (a)front side; (b)rear side [20.1].
664
20.4. PUBLIC EXPOSURE

The registrant or licensee is responsible for controlling public exposure
resulting from a nuclear medicine practice [20.6]. The presence of members of
the public in and near the nuclear medicine facility shall be considered when
designing the shielding and flow of persons in the facility. Exposure to members
of the general public from released patients also occurs, but this exposure is
almost always very small. The unintentional exposure of members of the public
in waiting rooms and on public transport is usually not high enough to require
special restrictions on nuclear medicine patients, except for those being treated
with radioiodine [20.3] who should receive patient-specific instructions for
limiting public exposure [20.8, 20.12]. In addition, exposure of those immediately
involved with the patient and the general population can occur through
environmental pathways including sewerage, discharges to water, incinerated
sludge or cremation of bodies. From the point of view of the individual doses
involved, this is of relativelyminor significance [20.1].
20.4.1. Visitors to patients
Arrangements should be made to control access of visitors (with special
emphasis on controlling access of pregnant visitors or children) to patients
undergoing radionuclide therapy and to provide adequate information and
instruction to these persons before they enter the patients room, so as to ensure
appropriate protection. Registrants and licensees should also take measures for
restricting public exposure to contamination in areas accessible to the public.
20.4.2. Radioactive waste
Registrants and licensees are responsible for ensuring that the optimization
process for measures to control the discharge of radioactive substances from a
source to the environment is subject to dose constraints established or approved
by the regulatory body [20.1320.15]. Chapter3 gives specific recommendations
for managing radioactive waste within the hospital facility.
For diagnostic patients, there is no need for collection of excreta and
ordinary toilets can be used. For therapy patients, there are very different policies
in different countries, but, in principle, the clearance criteria should follow a
dilution and decay methodology. Much of the activity initially administered is
eventually discharged to sewers. Storing a patients urine after therapy appears
to haveminimal benefit as radionuclides released into modern sewage systems
are likely to result in doses to sewer workers and the public that are well below
public dose limits [20.8]. Once a patient has been released from hospital, the
665
CHAPTER 20
excreted radioactivity levels are low enough to be discharged through the toilet in
their home without exceeding public dose limits. The guidelines given to patients
will protect their family, carers and neighbours, provided the patient follows
these guidelines.
20.5. RADIONUCLIDE THERAPY TREATMENT ROOMS AND WARDS
The following aims should be considered in the design of radionuclide
therapy treatment rooms and wards: optimizing the exposure to external radiation
and contamination, maintaining low radiation background levels to avoid
interference with imaging equipment, meeting pharmaceutical requirements, and
ensuring safety and security of sources (locking and control of access).
Typically, rooms for high activity patients should have separate toilet
and washing facilities. The design of safe and comfortable accommodation
for visitors is important. Floors and other surfaces should be covered with
smooth, continuous and non-absorbent surfaces that can be easily cleaned and
decontaminated. Secure areas should be provided with bins for the temporary
storage of linen and waste contaminated with radioactive substances.
20.5.1. Shielding for control of external dose
Radiation sources used in radiopharmaceutical therapy have the potential
to contribute significant doses to medical personnel and others who may spend
time within or adjacent to rooms that contain radiation sources. Meaningful
dose reduction and contamination control can be achieved through the use
of appropriate facility and room design. Shielding should be designed using
source related source constraints for staff and the public. The shielding should
be designed using the principles of optimization of protection and taking into
consideration the classification of the areas within it, the type of work to be done
and the radionuclides (and their activity) intended to be used. It is convenient to
shield the source, where possible, rather than the room or the person. Structural
shielding is, in general, not necessary for most of the areas of a nuclear medicine
department. However, the need for wall shielding should be assessed in the
design of a therapy ward to protect other patients and staff, and in the design of
rooms housing sensitive instruments (e.g. well counters and gamma cameras) to
keep a low background.
Special consideration should be given to avoiding interference with work
in adjoining areas, such as imaging or counting procedures, or where fogging of
films stored nearby can occur. Imaging rooms are usually not controlled areas.
666
Placing radiopharmaceutical therapy patients in unshielded hospital rooms

may expose persons in adjacent areas to levels that might cause dose constraints
to be exceeded. Vacating adjacent rooms or areas or installing shielding
(e.g. permanent poured concrete, solid concrete block, steel plates, lead sheets
or portable shielding devices) may be necessary to ensure dose constraints are
maintained in adjacent areas. Table20.1 gives typical shielding effectiveness
values for 131I. Exposure rate or dose rate measurements should be taken after
each radionuclide therapy administration, or worst case scenario evaluations
documented to confirm that these are below levels that could cause a dose
constraint to be exceeded.
Table20.1. TYPICAL SHIELDING EFFECTIVENESS VALUES FOR 131I
Material
Half-value layer
Tenth-value layer
Lead [20.16]
3.0mm
11mm
Concrete [20.17]
5.5cm
18cm
For permanent shielding evaluations, the design effective dose rate P

(in millisieverts per year or millisieverts per week) in a given occupied area is
derived by selecting a source related dose constraint, with the condition that
the individual effective doses from all relevant sources will be well below
the prescribed effective dose constraints for persons occupying the area to be
shielded. Table20.2 gives typical values for design effective dose in occupied
areas adjacent to a radionuclide therapy room [20.18]. A critical review of
conservative assumptions should be performed, so as to achieve a balanced
decision and avoid accumulation of over-conservative measures that may go far
beyond optimization.
It is preferable that patient treatment rooms be for individual patients and
adjacent to each other. If this is not possible, appropriate shielding between one
patient and another is required. When required, shielding should be provided for
nurses and visitors of radionuclide therapy patients, for which movable shields
may be used within patient rooms. When required, prior to each treatment,
movable shields should be placed close to the patients bed in such a way that
exposure of the nurses caring for the patient isminimized. This is achieved by
anticipating the nurses tasks, positions and movements throughout the room.
667
CHAPTER 20
Table20.2. TYPICAL VALUES FOR DESIGN EFFECTIVE DOSE p IN

OCCUPIED AREAS ADJACENT TO A RADIOTHERAPY TREATMENT
ROOM
Annual effective dose
(mSv/a)
Weekly effective dose

(mSv/week)
Occupational worker
10
0.2
Member of the public
0.5
0.01
20.5.2. Designing for control of contamination

Floors and other surfaces should be covered with smooth, continuous and
non-absorbent surfaces that can be easily cleaned and decontaminated. The floors
should be finished in an impermeable material which is washable and resistant to
chemical change, curved to the walls, with all joints sealed and glued to the floor.
The walls should be finished in a smooth and washable surface, for example,
painted with washable, non-porous paint.
Control of access is required to source storage, preparation areas and rooms
for hospitalized patients undergoing radionuclide therapy. A separate toilet room
for the exclusive use of therapy patients is recommended. A sign requesting
patients to flush the toilet well and wash their hands should be displayed to
ensure adequate dilution of excreted radioactive materials and tominimize
contamination. The facilities shall include a sink as a normal hygiene measure.
Bathrooms designated for use by nuclear medicine patients should be finished
in materials that are easily decontaminated. Hospital staff should not use patient
washing facilities, as it is likely that the floors, toilet seats and sink tap handles
will frequently be contaminated.
The design of safe and comfortable accommodation for visitors is
important. Shielding should be designed using source related dose constraints for
staff and the public. Secure areas should be provided with bins for the temporary
storage of linen and waste contaminated with radioactive substances.
20.6. OPERATING PROCEDURES
General advice on operating procedures in a nuclear medicine facility is
given in Chapter3. Management of radionuclide therapy patients should be
planned and performed in a way thatminimizes the spread of contamination in air
and on surfaces. Work with unsealed sources should be restricted to aminimum
number of locations.
668
20.6.1. Transport of therapy doses

Specific radiation safety considerations for the radiopharmacy are
addressed in Chapter9. Radiopharmaceuticals need to be transported within the
facility in shielded, spill-proof containers if warranted by the type of radionuclide
and amount of activity. The shielding should be such that external doses are
maintained as low as reasonably achievable (ALARA). The facility RPO should
be consulted in designing or evaluating the appropriateness of shielding and
transport methods.
20.6.2. Administration of therapeutic radiopharmaceuticals
Administration is normally by the oral route, intravenous injection
(systemic) or instillation of colloidal suspensions into closed body cavities
(intracavitary). Shielded syringes should be utilized during the intravenous
administration of radiopharmaceuticals as necessary to ensure that extremity
doses are maintained below occupational dose constraints. Absorbent materials or
pads should be placed underneath an injection or infusion site. The facility RPO
should be consulted to determine the necessity of other protective equipment (e.g.
shoe covers, step-off-pads, etc.) for particular radiopharmaceutical therapies.
For oral administrations of therapeutic radiopharmaceuticals, the
radioactive material should be placed in a shielded, spill-proof container. Care
should be taken tominimize the chance for splashing liquid or for dropping
capsules. Appropriate long-handled tools should be utilized when handling
unshielded radioactive materials. For intravenous administrations by bolus
injections, when dose rates warrant, the syringe should be placed within a syringe
shield (plastic for emitting radionuclides tominimize bremsstrahlung, high Z
materials for photon-emitting radionuclides) with a transparent window to allow
for visualization of the material in the syringe. For intravenous administrations
by slower drip or infusions, the activity container should be placed within a
suitable shield. For high energy photons, a significant thickness of lead or other
high Z material may need to be evaluated. In addition, consideration should be
given for shielding pumps and lines.
Procedures for administering a therapeutic radiopharmaceutical shall
include considerations to ensure as complete a delivery as possible of the
prescribed therapeutic activity. Any retention of material in syringes, tubing,
filters or other equipment utilized for administration should be analysed.
Where appropriate, equipment should be flushed or rinsed with isotonic saline
(or another physiological buffer) for parenteral administration or water for oral
administrations. All materials utilized in administrations shall be considered as
medical and radioactive waste, and should be labelled with the radionuclide, a
669
CHAPTER 20
radiation precaution sticker, and stored and or disposed of in a manner consistent

with local regulations.
20.6.3. Error prevention
Care should be exercised in avoiding administration of a therapeutic
radiopharmaceutical to the wrong patient. In addition, prior to administration, the
following should be verified:
The dose on the radiopharmaceutical label matches the prescription;
Identification of the patient by two independent means;
Identity of the radionuclide;
Identity of the radiopharmaceutical;
Total activity;
Date and time of administration;
Patients have been given information about their own safety.
The therapeutic radiopharmaceutical, activity, the date and time of
administration, and verification of the initial and residual assay should be entered
in some form in the patients medical record.
Pregnancy is a strong contraindication to unsealed radionuclide therapy,
unless the therapy is life-saving. This advice is all the more valid for radioiodine
therapy and for other radionuclides with the potential to impart radiation doses to
the fetus in the range of a few millisieverts. Therefore, where treatment is likely
or anticipated, the patient should be advised to take appropriate contraceptive
measures in the time prior to therapy [20.1]. Some radiopharmaceuticals,
including 131I as iodide and 32P as phosphate, rapidly cross the placenta, so that
the possibility of pregnancy should be carefully excluded before administration.
Before any procedure using ionizing radiation, it is important to determine
whether a female patient is pregnant. The feasibility and performance of
medical exposures during pregnancy require specific consideration owing to the
radiation sensitivity of the developing embryo/fetus [20.3]. Some procedures and
some radiopharmaceuticals (e.g. radioiodides) can pose increased risks to the
embryo/fetus. The ICRP has given detailed guidance in Publications 84 [20.19]
and 105 [20.2]. Radiation risks after prenatal radiation exposure are discussed in
detail in ICRP Publication 90 [20.20].
20.6.4. Exposure rates and postings
Values of ambient dose equivalent from the patient should be determined.
This information will assist in deriving appropriate arrangements for
670
entry by visitors and staff. Following the administration of the therapeutic

radiopharmaceutical to the patient, anterior exposure rates at the surface of and
1 m from the patient should be measured at the level of the patients umbilicus
(or other location as appropriate for the type of nuclear medicine administered),
using a calibrated radiation monitor (e.g. a portable ionization chamber).
Typically, these initial measurements are to be taken within 1h of administration
of the radiopharmaceutical therapy.
Rooms with radiotherapy patients should be controlled areas. A sign such
as that recommended by the International Organization for Standardization (ISO)
[20.21] should be posted on doors to the patients room and radioactive material
storage areas as an indicator of radiation (seeFig.20.2).
(a)
(b)
FIG.20.2. (a) International Organization for Standardization (ISO) radiation symbol;

(b)New IAEA/ISO radiation warning symbol.
It should be noted, however, that the ISO radiation symbol is not intended
to be a warning signal of danger but only of the existence of radioactive
material. A new symbol has been launched by the IAEA and the ISO to help
reduce needless deaths and serious injuries from accidental exposure to large
radioactive sources [20.22]. It will serve as a supplementary warning to the
trefoil, which has no intuitive meaning and little recognition beyond those
educated in its significance. The new symbol is intended for IAEA category 1, 2
and 3 sources [20.23] defined as dangerous sources capable of death or serious
injury, including food irradiators, teletherapy machines for cancer treatment and
industrial radiography units. The symbol is to be placed on the device housing
the source, as a warning not to dismantle the device or to get any closer. It
will not be visible under normal use, only if someone attempts to disassemble
the device. For radionuclide therapy applications, the new symbol will not be
located on building access doors, transport packages or containers. Rather, the
671
CHAPTER 20
ISO radiation symbol should be utilized to notify individuals of the existence of

radioactive material.
Facilities may also consider placing a radioactive precautions wristband
on the patients wrist if the patient is to remain in medical confinement. In
addition, for those patients remaining in medical confinement, the patient should
be resurveyed each day at the point of maximal uptake of the radiopharmaceutical.
The exposure rate or dose rate measured can then be used in determining the
activity remaining in the patient as well as developing appropriate release
instructions for the patient (seeSection 20.3).
20.6.5. Patient care in the treating facility
Medical practitioners should exercise their clinical duties consistent with
patient safety and good quality medical care. Unless otherwise specified by the
facility RPO, nurses, physicians and other health care personnel are to perform
all routine duties, including those requiring direct patient contact, in a normal
manner. However, medical practitioners should avoid lingering near the patient
unnecessarily and should spend as little time as necessary in close proximity to
radioactive materials or patients treated with radiopharmaceuticals and remain
at distances appropriate for the exposure rate or dose rate measurements from
such materials and patients. When necessary, portable shielding should be used to
reduce radiation levels to medical practitioners.
Ward nurses should be informed when a patient may pose a radioactive
hazard, and advice and training should be provided. The training should include
radiation protection and specific local rules, in particular, for situations where
there is a risk of significant contamination from, for example, urine, faeces or
vomiting. Appropriate training should also be given to night staff. In the case
of high activity patients, only essential nursing should be carried out. Other
nursing should be postponed for as long as possible after administration, to take
full advantage of the reduction of activity by decay and excretion. In addition,
there should beminimum handling of contaminated bed linen, clothing, towels,
crockery, etc. during the initial period and the instructions on how long these
precautions should be maintained should be documented.
The nursing staff should be familiar with the implications of the procedure,
the time and date of administration, and any relevant instructions to visitors.
Values of ambient dose equivalent at suitable distances should be determined.
This information will assist in deriving appropriate arrangements for entry by
visitors and staff. These arrangements should be made in writing in the local
rules.
672
20.6.6. Contamination control procedures

Work procedures should be formulated so as tominimize exposure from
external radiation and contamination, to prevent spillage from occurring and, in
the event of spillage, tominimize the spread of contamination. All manipulation
for dispensing radioactive materials should be carried out over a drip tray, in
order tominimize the spread of contamination due to breakages or spills.
Persons working with unsealed sources or nursing high activity patients
should wash their hands before leaving the work area. Patients treated with high
activity should use designated toilets. Simple precautions such as laying plastic
backed absorbent paper on the floor around the toilet bowl and instructions to
flush the toilet after each use will help tominimize exposure to external radiation
and contamination.
Particular attention and measures to limit spread of contamination are
required in the case of incontinent patients and, in cases of oral administration, if
there are reasons for believing that the patient may vomit. Contaminated bedding
and clothing should be changed promptly and retained for monitoring. Crockery
and cutlery may become contaminated. Local rules should specify washing up
and segregation procedures, except for disposable crockery and cutlery.
Where possible, a radionuclide therapy patient that requires medical
confinement should be placed in a private hospital room with a private toilet
and sink. The use of disposable plastic-backed absorbent pads or plastic sheeting
taped in place in the areas most likely to be contaminated, such as the floor around
the toilet and sink, may be appropriate for a facility. In all cases, consideration
of the ALARA principle should be maintained. Removal of loose contaminated
items from the patients room should be done on a daily basis.
In the event of a large volume spill of blood, urine or vomitus, medical
practitioners or staff should cover the spill with an absorbent material and
immediately contact the facility radiation safety service for appropriate cleanup
assistance and specific instructions. After such a spillage, the following actions
should be taken:
(a) The RPO should immediately be informed and directly supervise the
cleanup;
(b) Absorbent pads should be thrown over the spill to prevent further spread of
contamination;
(c) All people not involved in the spill should leave the area immediately;
(d) All people involved in the spill should be monitored for contamination
when leaving the room;
(e) If clothing is contaminated, it should be removed and placed in a plastic
bag labelled radioactive;
673
CHAPTER 20
(f) If contamination of skin occurs, the area should be washed immediately;

(g) If contamination of an eye occurs, it should be flushed with large quantities
of water.
Upon discharge and release of the patient, all remaining waste and
contaminated items should be removed and segregated into bags for disposable
items and launderable items. All radioactively contaminated waste items should
be labelled with the radionuclide and a radiation precaution sticker, and be
stored and or disposed of in a manner consistent with local regulations. The
patients room should be surveyed and checked for removable contamination
utilizing appropriate survey equipment (e.g. a GeigerMller counter or
scintillation survey meter). Where necessary, wipe tests should be performed.
Facility procedures should address applicable criteria for removable radioactive
contamination. Contamination monitoring is required for:
All working surfaces (including the interior of enclosures), tools,
equipment, the floor and any items removed from this area. Monitoring is
also required during the maintenance of contained workstations, ventilation
systems and drains.
Protective and personal clothing, and shoes, particularly when leaving an
area that is controlled due to the risk of contamination (monitors should be
available near the exit).
Clothing and bedding of therapy patients.
20.7. CHANGES IN MEDICAL STATUS
If the medical condition of a patient deteriorates such that intensive nursing
care becomes necessary, urgent medical care is a priority and should not be
delayed. However, the advice of the RPO should be sought immediately. In the
event of a deterioration in the patients medical condition, frequent or continual
monitoring of the patient may be necessary (e.g. septic shock, pulmonary
oedema, stroke or myocardial infarction). In some cases, the patient may need
to be transferred to intensive, special care or cardiac care units. It is possible
that patients in these units are in close proximity to each other with little or
no shielding available. As such, radionuclide therapy patients may present a
radiation hazard to other patients or medical practitioners. The nuclear medicine
physician and the RPO shall be notified of the transfer to a special unit as soon
as possible or prior to the transfer. The RPO shall determine whether portable
shielding is necessary to reduce doses to other patients or medical practitioners,
674
whether specific personnel monitoring is necessary, and whether specific

radiation precautions are necessary to keep radiation exposures ALARA.
20.7.1. Emergency medical procedures
Life-saving efforts shall take precedence over consideration of radiation
exposures received by medical personnel. This is particularly important for
therapy patients containing large amounts of radioactivity. Medical personnel
should, therefore, proceed with emergency care (e.g. when a patient has
suffered a stroke), while taking precautions against the spread of contamination
andminimizing external exposure. The staff should avoid direct contact with the
patients mouth, and all members of the emergency team should wear protective
gloves. Medical staff should be informed and trained on how to deal with
radioactive patients. Rehearsals of the procedures should be held periodically.
The only exceptional, life-saving situations are those medical emergencies
involving immediate care of patients in the case of strokes or similar situations,
when large amounts of radioactive material have been incorporated (of the order
of 2 GBq of 131I) [20.7].
20.7.2. The radioactive patient in the operating theatre
Radiation protection considerations should not prevent or delay life-saving
operations in the event that surgery on a patient is required. The following
precautions should be observed:
The operating room staff should be notified;
Operating procedures should be modified under the supervision of the RPO
tominimize exposure and the spread of contamination;
Protective equipment may be used as long as efficiency and speed are not
affected;
Rotation of personnel may be necessary if the surgical procedure is lengthy;
The RPO should monitor all individuals involved;
Doses to members of staff should be measured as required.
The RPO should consider whether personnel monitoring is required. The
number of persons in the operating theatre should beminimized, and operating
personnel should only remain in the operating room for theminimum amount
of time consistent with surgical objectives. If it is estimated that the circulating
blood or the area of the body to be treated surgically contains a significant
quantity of the radiopharmaceutical, the RPO and the surgeon should discuss
the procedures to be performed to keep radiation exposure to surgical personnel
675
CHAPTER 20
ALARA. The spread of radioactive contamination can beminimized through

the use of typical primary precautions used in the operating theatre. Radioactive
material can be kept off of surgeons through the use of gloves (the use of
double gloves may be appropriate). If an injury to surgical staff such as a cut or
puncture occurs, radioactive contamination of the skin or wound may occur. The
RPO should be consulted to evaluate contamination and any possible radiation
hazard, including the possibility of internal intakes. Any specimens sent for
pathological examination should be monitored for contamination. Tools and other
equipment from the surgery should be monitored for radioactive contamination,
decontaminated as necessary, and stored for radioactive decay or treated as
radioactive waste in accordance with local regulations.
20.7.3. Radioactive patients on dialysis
The care of patients receiving radiopharmaceutical therapy and who are on
dialysis may require additional consideration. In general, for systemic treatments,
these patients will not biologically clear radioactive materials as quickly as
typical patients since the clearance is highly dependent on the schedule of the
dialysis session. It may be necessary to reduce or otherwise adjust the activity
required for a therapy. The decision as to the activity required for such patients
should be based on either a trace trial administration of activity and the observed
elimination rate, or a careful review of the available literature for similar patient
administrations. Typically, the largest amount of radioactivity will be eliminated
during the first dialysis session following radiopharmaceutical therapy.
The RPO should assess the radiation exposures likely to be received
by medical practitioners during the sessions. In such cases, no significant
contamination of dialysis machines has been reported [20.1]. The materials,
tubing, filters and waste containers used during the sessions should be checked
by radiation safety staff to evaluate whether these need to be considered low level
radioactive waste and managed in accordance with facility and local regulations
(seeSection 20.4.2). There may be slight contamination of disposable items
such as liners and waste bags, which may require storage for some time, in the
case of 131I. In most cases, however, no special precautions will be required and
the dialysis and radiation safety staff will advise patients on how to deal with
disposables.
20.7.4. Re-admission of patients to the treating institution
If a patient who still contains a therapeutic amount of radioactive material
is re-admitted to the treating institution, the RPO shall be notified as soon as
possible after re-admission. Patient medical charts should include information on
676
dates of cessation of radiation precautions (perhaps in electronic chart systems

which could provide useful triggers for the needed precautions in the event
of re-admission). The RPO shall monitor the patient and specify any required
precautions to be followed by medical practitioners. Where required, radiation
precaution tags should be placed on the patient, the patients room and chart.
20.7.5. Transfer to another health care facility
Some patients may need to be transferred to another health care facility
(i.e. another hospital, skilled nursing facility, nursing home or hospice, etc.)
following therapy treatments. In such a case, care must be taken that, in addition
to practical measures and advice to ensure safety of other staff, any legal
requirements relevant to the second institution are also complied with [20.1].
Patients transferred to another health care facility should meet the criteria for
unrestricted clearance. However, the possibility for the generation of low level
radioactive waste should be examined by the RPO of the treating facility and any
issues should be discussed with the facility accepting the patient transfer. In the
rare event that a patient being transferred to another health care facility does not
meet the criteria for unrestricted clearance, the RPO shall ensure that the facility
accepting the patient transfer has an appropriate registration or licence that would
allow acceptance of the patient with therapeutic amounts of radioactive materials
on board. The RPO should provide radiation safety information and precautions,
if any, for the patient and for the receiving health care facility.
20.8. DEATH OF THE PATIENT
Therapeutic amounts of radioactive materials are typically not administered
to critically ill patients unless there are circumstances where the palliative
use of radioactive materials in terminal patients will significantly improve
the quality of life of the patient. However, should the patient die in the period
immediately following therapy, special consideration may need to be given to
the treatment of the corpse. To facilitate this, the patient should be given a small
card with details of their treatment and contact details for a radiation protection
specialist/medical physicist associated with the department responsible for
the therapy (seeFig.20.1). Additional specific guidance for the death of a
radionuclide therapy patient has been developed by the IAEA [20.1].
Areas of concern arise with respect to embalming, burial or cremation of
the corpse and the conduct of autopsy examinations. National regulations, some
quite dated, are available for some or all of these in many countries, but there is a
lack of international recommendations. Practice tends to be guided by an untidy
677
CHAPTER 20
mixture of custom, professional guidance and national regulation [20.1]. Recent

reports have emphasized the need to be sensitive to the wishes of the deceased
and their family when decisions about the disposal of the corpse are being made.
This may be particularly important if the possibility of retaining some organs for
radiation protection reasons is being considered [20.1].
The authorities in many countries now place limits on the radioactivity that
may be present in the corpse before autopsy, embalming, burial or cremation.
No special precautions are required for direct burial or cremation, without
embalming, provided the activity involved is not in excess of national limits.
No special precautions are required for embalming if activities do not exceed
the levels mentioned in table 14 of Ref.[20.1] for autopsy. If the activities are
greater, then a corpse should not normally be embalmed, but if embalming is
required an RPO should be consulted.
20.8.1. Death of the patient following radionuclide therapy
In cases where the death occurs in a hospital, access to the room occupied
by the deceased should be controlled until the room has been decontaminated
and surveyed. Radioactive bodies should be identified as potential hazards by
a specified form of identifier. Identification of the possibility that a body may
contain radioactive substances relies on information provided in the patient
records, the information card (Fig.20.1) or information gleaned from relatives
or others. A body bag may need to be used to contain leakage of radioactive
substances. Tominimize external radiation risk, the corpse may need to be
retained in a controlled area.
In the event that a patient dies within the treating health care facility while
still containing a therapeutic quantity of radioactive material, the treating medical
practitioner and the RPO shall be notified immediately. Depending on the number
of days that have elapsed between radiopharmaceutical treatment and death, the
radiation hazard may have been reduced considerably, and precautionsminimized.
In the rare event that large quantities of radiopharmaceuticals are still within
the body, the RPO shall identify specific radiation precautions as necessary,
depending on the type of radionuclide and measured exposure rates or dose
rates. Nursing staff should be provided with instructions informing them that
the normal procedure of pressing down on the abdomen of a corpse must not be
performed due to the radiation and/or contamination levels that may result [20.1].
The RPO shall notify the morgue prior to the arrival of the body, and the RPO
should discuss radiation safety precautions with morgue personnel, as required.
In most cases, if the patient has already been released from the treating
facility, no special precautions are generally necessary for handling the body.
678
20.8.2. Organ donation

If organ donation is being considered, the RPO shall determine necessary
precautions for operating theatre personnel who will harvest the organ(s). Unless
the organ is directly involved in the treatment regime, it is unlikely that the
donated organ will contain an amount of radioactive material to cause significant
damage to the organ or deliver a radiation dose to the recipient sufficient to
nullify the donation. However, the nuclear medicine physician and RPO should
be prepared to estimate such quantities and doses.
20.8.3. Precautions during autopsy
The dose constraints applying to pathology staff responsible for the conduct
of autopsy examinations will be either those for the general public or those
for radiation workers, depending on the training and classification of the staff
concerned. However, it is almost inevitable that some members of the pathology
staff will be classified as members of the public from a radiation protection point
of view. These constraints and the radiation safety procedures to be applied
in practice should be determined in close consultation with the RPO from the
department in which the therapy was administered. Where the possibility that the
corpse may be radioactive arises, a proposed autopsy should be suspended until
the situation is clarified to the greatest extent possible and a risk assessment has
been undertaken by the RPO. This should establish the type, nature and location
of the radioactive material used and when the therapy occurred. Any reporting
or notifications required by law and/or good practice should also be undertaken.
Additional points to consider for autopsies of radioactive bodies are noted in
annex IV of Ref.[20.1].
Although it is rare that the body of a patient will be sent for autopsy shortly
after administration of a therapeutic radiopharmaceutical, if death occurs within
2448h post-administration, a considerable amount of activity may be present in
blood and urine. In these cases, the RPO or radiation safety staff should supervise
the autopsy. Any residual activity in tissue samples should be evaluated prior to
releasing the samples to the pathology laboratory. If death occurred more than
48h post-administration, there will typically be little, if any, activity in the blood
or urine. In these cases, activity may only be present in residual treated areas or
metastatic disease sites. The staff dose may be reduced by deferring the autopsy
where necessary and practical. Finally, Singleton et al. [20.24] conclude that:
provided that appropriate precautions are implemented, determined
through consultation with a qualified expert in radiation protection and by
679
CHAPTER 20
completion of risk assessment, the radioactive autopsy can be undertaken

safely and in compliance with relevant legislative requirements.
Unsealed radioactive substances may be present in a particular body cavity
or organ, or they may have concentrated after systemic administration (e.g. 131I
in the thyroid gland). Drainage of the cavity or excision of the organ will reduce
exposure if undertaken at the start of the autopsy. In addition, care should be
given with respect to organs with significant activity. In cases where the patient
had received a dose of emitting colloid or spheres (e.g. 32P chromic phosphate
into a body cavity or 90Y microspheres into the liver), significant activity may be
present in the cavity fluid or in the embolized organ. Beta radiation sources may
provide a significant dose to the hands because they will be in close contact with
body tissues and fluids [20.5]. Autopsy and pathology staff should wear standard
protective clothing (i.e. gloves, lab coats, eye protection, etc.) and personnel
monitoring should be considered. For emitters, double surgical gloves may be
helpful in reducing skin exposures. An intake of airborne material inadvertently
released during cutting or movement of radioactive tissue or organs can be
prevented by wearing eye protection and a face mask.
20.8.4. Preparation for burial and visitation
Funeral directors will need to be advised of any necessary precautions, and
notification of the relevant national competent authorities may be required. It is
essential that funeral directors andministers of religion do not overreact to the
risks associated with the radioactive corpse. Careful communication is needed
to ensure that adequate controls are implemented without compromising dignity.
Situations where the wishes of the next of kin have to be significantly disrupted
should be rare [20.1].
In cases where the body will be prepared for burial without autopsy or
embalming, if the RPO believes that the potential dose likely to be received by
the personnel preparing the body will not exceed the appropriate dose constraint,
burial can proceed. In rare cases where dose constraints may be exceeded, the
RPO should provide radiation precaution information (e.g. restricting the time
spent near the body). In cases where the body will be prepared for burial by
embalming, the RPO should notify the morgue or funeral home that the body
contains therapeutic quantities of radioactive material and should provide them
with precautions tominimize radiation exposure and radioactive contamination.
Embalming is conducted by injecting an embalming fluid into the body and
flushing body fluids into the drain. Embalming staff should wear standard
protective clothing (i.e. gloves, lab coats, eye protection, etc.) and personnel
680
monitoring should be considered. Careful cleaning of equipment in the usual

manner will remove radioactive contamination [20.5].
In most cases, no precautions will be necessary during visitation. If the
possibility exists that there are measurable dose rates at 30cm from the body, the
family and funeral home should be given appropriate precautions, as necessary to
provide assurance that dose constraints are met.
20.8.5. Cremation
A proportion of the activity retained will appear in cremated remains and
may be sufficient, particularly in the case of long lived radionuclides, to require
controls to be specified. The main concern is in respect to the scattering of ashes,
although contact dose rates with the container may have to be considered if
cremation takes place shortly after administration [20.1].
Assuming that the body has been prepared in accordance with the
recommendations in the preceding sections, no additional handling precautions
are necessary in transporting the body to the crematorium. The crematorium
personnel should be informed by the treating facility or family that the body might
contain radioactive material. Crematorium personnel may contact the treating
facility if they need additional guidance on handling the body. Such guidance
should include methods tominimize radiation exposure, contamination of the
retort and especially methods tominimize radioactive ash particles. Crematorium
employees may receive external exposure from the radioactive body or from
contamination of the crematorium or internal exposure from inhalation of
radioactive particles while handling the ashes [20.25]. Bodies that contain
emitting radionuclides may result in some external exposure to employees of
the crematorium. No precautions are necessary as long as there isminimal time
required to handle the body at the crematorium (a likely assumption). Cremation
of non-volatile radionuclides might result in contamination of the retort. As the
most significant hazard from this contamination is inhalation of ash particles
during cleaning of the retort, it is appropriate for workers who clean the retort to
wear dust masks and protective garments.
The most likely hazard to the general population in the vicinity of the
crematorium is the inhalation of radioactive material emitted with the stack
gases. Each crematorium should maintain records of the type and activity in
bodies cremated, when known.
The potential for effective doses from cremation of bodies containing
131
I should be evaluated and some have suggested that if a crematorium were
to handle bodies that contain 131I and do not exceed 100 GBq in a single year,
the effective dose to individuals in the surrounding population would not likely
exceed 0.1 mSv [20.5]. It, therefore, appears that no specific radiation hazard
681
CHAPTER 20
would exist even if a crematorium were to handle several bodies per year
containing 131I.
REFERENCES
[20.1] INTERNATIONAL ATOMIC ENERGY AGENCY, Release of Patients After
Radionuclide Therapy, Safety Reports Series No. 63, IAEA, Vienna (2009).
Radiological Protection in Medicine, Publication 105, Elsevier, Oxford (2008).
Recommendations of the ICRP, Publication 103, ICRP, Elsevier, Oxford (2008).
[20.4] NUCLEAR REGULATORY COMMISSION, Consolidated Guidance about Materials
Licensees, Rep. NUREG-1556, Vol. 9, Office of Standards Development, Washington,
DC (1998).
[20.5] NATIONAL COUNCIL ON RADIATION PROTECTION AND MEASUREMENT,
Management of Radionuclide Therapy Patients, Rep. No. 155, Bethesda, MD (2006).
[20.6] ZANZONICO, P.B., SIEGEL, J.A., ST. GERMAIN, J., A generalized algorithm for
determining the time of release and the duration of post-release radiation precautions
following radionuclide therapy, Health Phys. 78 (2000) 648659.
[20.7] INTERNATIONAL ATOMIC ENERGY AGENCY, Applying Radiation Safety
Standards in Nuclear Medicine, Safety Reports Series No. 40, IAEA, Vienna (2005).
[20.8] INTERNATIONAL COMMISSION ON RADIOLOGICAL PROTECTION, Release
of Patients After Therapy with Unsealed Sources, Publication 94, Pergamon Press,
Oxford (2004).
[20.9] STRAUSS, J., BARBIERI, R.L. (Eds), Yen and Jaffes Reproductive Endocrinology,
6th edn, Saunders, Elsevier, Philadelphia, PA (2009).
[20.10] DAUER, L.T., WILLIAMSON, M.J., ST. GERMAIN, J., STRAUSS, H.W., Tl-201
stress tests and homeland security, J. Nucl. Cardiol. 14 (2007) 582588.
[20.11] DAUER, L.T., STRAUSS, H.W., ST. GERMAIN, J., Responding to nuclear granny,
J. Nucl. Cardiol. 14 (2007) 904905.
Radiological Protection in Medicine, Publication 73, Pergamon Press, Oxford (1996).
[20.13] INTERNATIONAL ATOMIC ENERGY AGENCY, Applications of the Concepts of
Exclusion, Exemption and Clearance, IAEA Safety Standards Series No. RS-G-1.7,
IAEA, Vienna (2004).
[20.14] INTERNATIONAL ATOMIC ENERGY AGENCY, Regulatory Control of Radioactive
Discharges to the Environment, IAEA Safety Standards Series No. WS-G-2.3, IAEA,
Vienna (2000).
682
[20.15] INTERNATIONAL ATOMIC ENERGY AGENCY, Management of Waste from

the Use of Radioactive Material in Medicine, Industry, Agriculture, Research and
Education, IAEA Safety Standards Series No. WS-G-2.7, IAEA, Vienna (2005).
[20.16] DELACROIX, D., GUERRE, J.P., LEBLANC, P., HICKMAN, C., Radionuclide and
Radiation Protection Data Handbook, Oxford University Press, Oxford (2002).
[20.17] SCHLEIEN, B., BIRKY, B., SLABACK, L., Handbook of Health Physics and
Radiological Health, 3rd edn, Williams and Wilkins, Baltimore, MD (1998).
[20.18] PODGORSAK, E.B. (Ed.), Radiation Oncology Physics: A Handbook for Teachers
and Students, IAEA, Vienna (2005).
Pregnancy and Medical Radiation, Publication 84, Pergamon Press, Oxford (2000).
Biological Effects after Prenatal Irradiation (Embryo and Fetus), Publication 90,
Pergamon Press, Oxford (2003).
[20.21] INTERNATIONAL ORGANIZATION FOR STANDARDIZATION, Basic Ionizing
Radiation Symbol, ISO 361, Geneva (1975).
[20.22] INTERNATIONAL ORGANIZATION FOR STANDARDIZATION, Ionizingradiation Warning Supplementary Symbol, ISO 21482, Geneva (2007).
[20.23] INTERNATIONAL ATOMIC ENERGY AGENCY, Categorization of Radioactive
Sources, IAEA Safety Standards Series No. RS-G-1.9, IAEA, Vienna (2005).
[20.24] SINGLETON, M., START, R.D., TINDALE, W., RICHARDSON, C., CONWAY, M.,
The radioactive autopsy: safe working practices, Histopathology 51 (2007) 289304.
[20.25] WALLACE, A.B., BUSH, V., Management and autopsy of a radioactive cadaver,
Australas. Phys. Eng. Sci. Med. 14 (1991) 119124.
683
Appendix I
ARTEFACTS AND TROUBLESHOOTING
E. BUSEMANN SOKOLE
Academic Medical Center,
Amsterdam, Netherlands
N.J. FORWOOD
Royal North Shore Hospital,
Sydney, Australia
I.1. THE ART OF TROUBLESHOOTING
I.1.1. Basics
Troubleshooting refers to the process of recognizing and identifying the
cause of an artefact, a malfunction or a problem in an instrument. The problem
could be immediately obvious, for example, the instrument does not work at all
or a particular component stops working (such as the computer, the mechanism
for whole body scanning or the automatic mechanism for collimator exchange).
The malfunction could also be less obvious, and be recognized only by an
abnormality in the expected result (such as the pattern formed by a defective
photomultiplier tube (PMT) in the gamma camera clinical or quality control
(QC) image or an unexpected calibration result in a radionuclide dose calibrator).
Such an abnormality is generally referred to as an artefact, in particular, when
observed in images.
The malfunctioning of an instrument can occur at any time. It might
become evident from a routine QC test. However, it is especially stressful when
it occurs during a patient investigation. In such a situation, the first lines of action
are tominimize the distress to the patient that a problem has occurred, to remain
calm and clear headed, to immediately try to identify the problem and correct it,
if possible, and to decide whether the investigation can be continued, either on
the same instrument or another similar one, or whether the investigation must be
rescheduled. An action flow chart is useful in the decision making process. Such
a flow chart is shown in Fig.I.1 for actions following a QC test.
684
- recognizing
FIG.I.1. Decision tree suggested for performance, evaluation and follow-up of a quality
control test. The symbols indicate: a start or end; b process to be performed; c
protocol; d intermediate results; e checks required; f decision to be made; g action
taken. Question answer: Y yes; N no.
Good communication and teamwork between the personnel of the

department are essential, especially when the consequences of an instrument
failure or malfunction may involve the action of different disciplines (e.g. taking
care of the patient, undertaking first line troubleshooting and problem solving,
685
APPENDIX I
decision making if the problem cannot be immediately solved, patient

rescheduling).
Available qualified personnel who understand the basic functioning
of the instrumentation and the digital environment (e.g. the imaging or
measuring instrument, computers, peripherals, network, picture archiving and
communication system) are desirable. Up to date protocols for instrumentation
function and set-up, instrument calibration, work procedures, clinical studies,
QC tests (including action thresholds) and phantom preparation are necessary for
creating and maintaining uniform methods within the department. These are an
important reference during troubleshooting.
Illustrative examples of artefacts shown in this appendix are restricted to
only a few examples. The IAEA Quality Control Atlas for Scintillation Camera
Systems is a further valuable resource with many more examples of gamma
camera artefacts.
I.1.2. Processes in troubleshooting
Troubleshooting in a clinical environment requires first and foremost
immediate action, clear thinking and resources to a network of qualified personnel,
in order tominimize down time. The following processes are suggestions to assist
when setting up a troubleshooting system within the department:
(a) Identify a qualified person within the department who will have the
responsibility for that day to be called upon as first-line support should a
problem occur. This person could be the physicist, technologist or technical
engineer. They should be called upon immediately when a problem is
signalled and be responsible for communication regarding the problem and
decisions made.
(b) When a problem is signalled during a patients nuclear medicine
investigation, make every effort to localize the problem as soon as
possible. Decide whether the problem can be solved immediately, so that
the investigation can be continued (possibly without moving the patient
beneath an imaging system), or whether the problem solving will take more
time and the patient has to leave the room and return to the waiting room
until further decisions have been made.
(i) Example A: An example of necessary fast action is when the computer
or the computer network halts during acquisition of a planar dynamic
gamma camera study started immediately after injection of the
radiopharmaceutical (e.g. renography). If at all possible, the problem
should be solved immediately and the dynamic study continued or
restarted. Depending on the type of study, the diagnostic value of the
686
study may still be salvaged, without requiring the patient to return on

another day and receive another radioactive injection. The problem of
patching the dynamic study between the first and second parts with
missing data may then be tackled afterwards. The nuclear medicine
physician should decide whether the interrupted study still has
diagnostic value.
(ii) Example B: A problem during an electrocardiogram (ECG) gated
cardiac study may be related to an inappropriate ECG signal to the
computer, simply requiring the repositioning of the ECG leads (e.g. a
negative R-wave instead of the correct positive R-wave).
Caution: If the data acquisition computer and/or imaging system is to be
shut down and restarted, the patient must first be removed from the patient
pallet.
(c) If the identity of the problem is not obvious or the problem cannot be
solved immediately, a decision must be made as to whether the instrument
is totally unusable or usable with limitations until repaired. A partially
performed clinical investigation should, if possible and appropriate, be
repeated on another similar instrument in the department. This also applies
to the other investigations scheduled for that instrument for patients already
administered with radioactivity. Any change in instrumentation or protocol
must also be noted in the patient record.
Caution: Caution should be exercised regarding the comparative validity
of quantitative data when a study is performed on another instrument
(e.g. cardiac studies assessing left ventricular ejection fraction).
(d) The problem may need to be solved by an in-house service, a telephone
consultation with the service centre of the vendor or by a vendors
service visit, which should be initiated as soon as possible. To assist
with localizing the problem, as much information as possible should be
documented regarding the circumstances at the time of malfunction, such
as other activities being performed (e.g. data processing, data transfer over
a computer network, temperature and humidity, power stability, nearby
surrounding activities, time of day). A digital photograph of the situation
and any error message display on the instrument monitors can be a helpful
tool for troubleshooting. An example from a digital log book (created in
house using File Maker Pro software) is shown in Fig.I.2.
687
APPENDIX I
FIG.I.2. Example of the digital documentation (in Dutch, with log translation) of a gamma
camera gantry error. This is one item from the digital database (developed in house) of
a troubleshooting log. The error report includes the instrument type, the date of problem
report, action priority, room location, name of responsible person, company information, log
describing the problem, first-line actions taken and their results. A photograph is included of
the gantry error message readout.
(e) Enter all problems and as much related data as possible into the log book
specific for the instrument. The solutions should also be documented. A
well documented and maintained digital record can be especially useful for
assisting with troubleshooting by a search for a previous similar problem or
if a repeat problem occurs at a later date. This log book should be started
at installation and maintained throughout the lifetime of the instrument,
together with preventive maintenance reports and any major modification.
Such a log book can also be linked to the QC results and records.
688
(f)
A problem may manifest itself during a routine QC test (such as an artefact

from a malfunctioning PMT observed in a routine QC uniformity image).
A calibration procedure may fail or show values that are outside the
acceptable range. A decision must then immediately be made regarding the
acceptability of continuing to use that instrument, whether the instrument
can be used with recognized limitations, or whether it must be taken
out of use until the problem is solved. This decision making should be
communicated with the other responsible staff members. The problem
and follow-up actions must be documented in the log book. Two examples
of artefacts discovered at the time of routine QC testing of flood-field
uniformity are shown in FigsI.3 and I.4.
At the moment of discovering a problem in a QC test, it is uncertain when
the malfunction causing the artefact or calibration failure first occurred.
The assumption is that it may have occurred at any time between the
current and previous QC test. The clinical studies prior to the current QC
test should, therefore, be carefully reviewed in order to ascertain when
the artefact first occurred, and if the artefact in the clinical images might
FIG.I.3. Weekly system uniformity image (left) from one detector of a dual head gamma
camera. The image was obtained with all corrections activated (linearity, energy, uniformity),
low energy high resolution collimator, 57Co flood source, symmetric energy window over
122 keV, 256 256 matrix, 4 million counts. On the lower left side, there is an irregular
hot semicircular area. The National Electrical Manufacturers Association (NEMA) uniformity
quantification in the useful field of view (UFOV; right image outer rectangle) confirms that
this non-uniformity is outside of specifications. The problem was a loss of gel between the
border photomultiplier tube and crystal. Once the gel was replaced, the uniformity was
restored. Note: This camera required a service. The defect affected imaging at the edge of the
field of view. This detector could, therefore, still be used for imaging within the central field of
view (CFOV; right image, inner rectangle) area, for which the NEMA differential uniformity
values were satisfactory (planar and whole body imaging, and with caution SPECT imaging).
689
APPENDIX I
FIG.I.4. Routine system uniformity images from a dual head gamma camera. The images
were obtained with all corrections activated, low energy all-purpose collimators, 57Co flood
source, symmetric energy window over 122keV, 6464 matrix, 16 Mcounts. The uniformity
was quantified with the National Electrical Manufacturers Association (NEMA) integral and
differential uniformity parameters in both the useful field of view (UFOV) and central field of
view (CFOV). The values from detector head 1 were within the expected limits. Detector head 2
shows a gross non-uniformity pattern corresponding to the photomultipliers. This pattern was
due to a failure of the electronic correction due to bad electrical contacts of the circuit boards.
After re-seating the relevant circuit boards, the problem was solved and uniformity was
restored as shown in a follow-up test (not shown here). Note: The non-uniformity of head 2
was extensive and, thus, imaging with this detector had to be suspended until the problem was
solved.
have resulted in an incorrect diagnostic report. If it appears that the artefact

may have compromised the images and report, a decision must be made
whether to recall the patient and redo the study after the problem has been
solved, or redo the study on another instrument. An example of an artefact
not discovered until the following QC test is shown in Fig.I.5. The nuclear
medicine physician should be informed and consulted.
(g) Particular care must be exercised at all times to be alert to artefacts in
clinical images, abnormal quantitative readings and data analysis results.
An obvious problem is present when an organ uptake measurement is
>100%. Constant alertness is an ongoing process, which should be an
integral part of daily practice for all members of the nuclear medicine
team.
If the same or a very similar abnormal pattern is observed in successive
clinical images from different patients, then the abnormal pattern may
690
be caused by a malfunction in the instrument rather than metabolic

dysfunction in the patients. If such a situation is suspected, the problem
should first be investigated before further patients are injected and imaged.
Troubleshooting may involve not only investigating the instrument, but also
checking the radiopharmaceutical quality and integrity of the radioactive
administration, etc. It often requires a QC test to assess the situation. An
example of uniformity artefacts in lung perfusion studies that were not
immediately related to instrument artefacts is shown in Fig.I.6.
The acquired image data for an investigation should always be reviewed
carefully before the patient is allowed to leave the department. An artefact
or inadequate data may require that the data acquisition be repeated.
FIG.I.5. A 99mTc phosphonate bone scan obtained with a dual head gamma camera.
(a)Anterior and posterior whole body images. (b)R lateral and L lateral static images of the
left knee. (c)Routine system uniformity quality control image of detector 1 taken 2 d later. The
photomultiplier tube artefact is at the upper border of the field of view. Note: The bone scan
was reported without noticing the malfunctioning photomultiplier tube of detector 1, which was
only discovered at the following routine QC test. On review, the effect of the photomultiplier
tube artefact was not discernible in the anterior whole body scan made with detector 1 (a), but
was visible in the R lateral static of the bone scan using a colour table and high contrast that
highlighted low count areas. This example illustrates alertness to an unexpected malfunction.
This camera required a service, but could still be used with caution for planar imaging within
a limited part of the detector. Owing to the nature of the clinical bone study and the location
of the photomultiplier tube artefact, this study was not repeated. The gamma camera required
a service.
691
APPENDIX I
FIG.I.6. A clinical lung perfusion study using 99mTc macroaggegates. (a)Images of the lungs
were obtained with camera 1 (top row images: posterior, right posterior oblique, right lateral;
bottom row images: anterior, left lateral, left posterior oblique). The irregular pattern of hot
and cold areas was not at first recognized as a camera problem. After two subsequent patients
demonstrated the same patchy pattern in their lung perfusion images, the clinicians reviewing
the studies questioned the results and troubleshooting was initiated. Quality control of the
radiopharmaceutical was acceptable. A uniformity quality control test of the gamma camera
was made and this revealed gross non-uniformity ((c) left image). All patients were recalled
and re-imaged on camera 2. (b)Lung perfusion images obtained on camera 2 (with the same
image order as in (a)). Camera 1 was retuned, which restored uniformity ((c) right image).
Further investigation revealed that there had been a power disruption during the night. This
had corrupted the energy correction values, and explained the reason for the non-uniformity.
Note: A routine quality control uniformity test had not been performed at the start of the days
clinical imaging. If this had been done, the problem would have been identified immediately.
An uninterruptible power supply was later installed in order to prevent a similar future
occurrence. (For more details, see the IAEA Quality Control Atlas for Scintillation Camera
Systems.)
(h) After a problem has been solved, the instrument should be tested for correct
functioning before being released for clinical use. If computer software or
hardware has been changed, reboot and restart the system to ensure that the
system works after a power down:
(i) Be aware of any changes in hardware or software that could affect
quantitative results. Validate the results.
692
(i)
(ii) Changes to hardware may require QC testing before the instrument is

released for clinical use.
Be aware of an intermittent or repetitive problem. Creative testing and
dedicated persistence is required to locate the cause of such a problem.
Even after a problem appears to be solved, it may still be present because
of instability in a component. For example, this has been the case with
electronic grounding, cable connections, cable breaks and a fluctuating
power supply. An example of the effect of voltage instability in a single
photon emission computed tomography (SPECT) study is shown in Fig.I.7.
Continued alertness is always required, as well as repeated QC testing.
I.1.3. Troubleshooting remedies

Various first-line troubleshooting tactics can be useful before resorting
to contacting the service. If a service contract is available for an instrument, it
should be clear where the responsibilities and limits lie. Some general hints to be
considered are given below, although the circumstances are left to the discretion
of the troubleshooter. The troubleshooting section of the instruction manual of
the instrument should also be consulted.
In the event of failure or instability of an instrument or component (to be
carried out by a qualified person or the service engineer):
(a) Check electrical power, circuit breakers, fuses, cables and cable
connections, fans;
(b) For accessible batteries, check the level of battery power within the
instrument that regulates a specific function;
(c) Check for dust, and cleanliness of sensors and metal contacts.
Computer and network:
(a) If a program halts, exit and restart the program. If this is unsuccessful, shut
down and restart the computer. (If restarting the data acquisition computer,
make sure that the patient is not on the imaging table.)
(b) For a suspected communications failure between the instrument and the
computer, shut down and restart the computer. If this does not solve the
problem, shut down both the computer and the instrument, and, after about
30 s, restart the instrument and then the computer. Be careful to follow
a correct startup procedure and make sure that the patient is not on the
imaging table.
(c) If peripheral equipment (such as a printer) stops functioning or produces an
error message, shut down and restart that equipment.
693
APPENDIX I
FIG.I.7. (a)Quality control images over the whole field of view of the acquisition data of a
SPECT myocardial perfusion study (left one projection image, middle sinogram over the
whole field of view (X), right linogram over the whole field of view (Y)). The images were
obtained from a 3-detector SPECT system (120 rotation per head, starting with head 1, and a
360 total rotation). The linogram shows an upwards shift in the images from head 1 towards
the finish of the 120 rotation (first third of the dataset). In order to clarify the situation, a
point source was placed off-axis and imaged with the same data acquisition parameters. (b)In
order to test the system, a SPECT acquisition was made of a point source placed off-axis.
Quality control images of this acquisition (same image order as above), and their quantitative
offset analysis (lower row). Offsets are seen in both X and Y in detector 1 data, identified
clearly by the jump in offset in both X and Y on the quantitative analysis. The problem was due
to a decrease in voltage to the signal board of detector head 1 at certain projection angles.
This was found to be due to instability in the power cable connected to the signal board.
The problem was resolved only after replacing the cable. Note: This problem was difficult to
locate. A problem was signalled in patient studies by the reporting nuclear medicine physician
who observed upwards motion in the quality control review of the patient SPECT data from
successive patients. Initially, a corrupted centre of rotation calibration was considered to be
the cause. However, the problem repeated itself, and was again recognized on subsequent
clinical and point source SPECT acquisitions. It took much persistence from the department
to keep on testing the system and several visits by the service engineer before the problem was
found. The upwards shift was only seen in one detector head, thus pointing to a problem with
the camera and not movement of the patient, which would have been seen in the acquired data
of all three heads, at the same time frames.
694
(d) Check the computer network and communications.

(e) Note: Before instrument re-use, a simple QC check may be necessary to
ensure that the instrument is functioning correctly.
Error and artefact in results:
(a) Check the radiopharmaceutical and injection quality, study parameters and
instrument settings, and patient positioning before investigating instrument
malfunction. Consult the department procedure manuals.
(b) If observed in a QC test, check that the test method was correct.
(c) Initiate appropriate supplementary QC tests, as appropriate.
I.2. IMAGE ARTEFACTS
I.2.1. Recognizing image artefacts and their underlying causes
Pattern recognition is the essential ingredient of interpretation of nuclear
medicine images. Thus, recognizing an artefact is also an essential part of pattern
recognition. Image artefacts manifest themselves in different ways as a result
of different factors. Relating an artefact to the underlying problem causing the
artefact is a developing process of understanding the instrument and how it
should be used. A particular instrument may show characteristic artefact patterns
that repeat and become familiar over time.
An artefact may also be caused by incorrect instrument settings or be patient
related. Thus, a daily component of troubleshooting is to review the acquired
data before each patient is allowed to leave the department.
Some causes of problems encountered in gamma camera images are given
below for different performance parameters. Related images can be found in
the IAEA Quality Control Atlas for Scintillation Camera Systems, which is an
extensive resource of different types of image artefact that may be encountered
in planar, whole body and SPECT imaging modes of a gamma camera system.
Examples given in the Atlas include results from QC tests as well as clinical
examples.
Problems encountered in gamma camera images for different performance
parameters:
(a) Distortion of the energy spectrum photopeak shape, loss of energy
resolution: A change may be caused by:
(i) Poor tuning or energy calibration.
(ii) Malfunctioning PMT or preamplifier.
695
APPENDIX I
(iii) Inadequate or unstable electrical grounding.

(iv) Instability in electrical contact in the detector power supply.
(v) Deteriorating detector material.
(vi) Interference from nearby radionuclides.
(b) Decrease in detector sensitivity: The decrease in count response may be
related to:
(i) Incorrect centering of the photopeak window.
(ii) Change in the PMT tuning values or gain values.
(iii) Malfunctioning PMT.
(iv) Deteriorating detector material.
(c) Poor image uniformity of a gamma camera: Image uniformity may be
affected by a variety of problems, for example:
(i) Deterioration in detector properties, so that energy and/or linearity
corrections no longer correspond.
(ii) Inadequate energy correction for radionuclides other than 99mTc.
(iii) Offset in centering of the image with respect to the image matrix and
corrections.
(iv) Poor tuning of the PMTs.
(v) Malfunctioning or defective PMT(s).
(vi) Loss of optical coupling between PMT and light guide, light guide and
crystal surface, or PMT and crystal surface.
(vii) Asymmetrical or erroneous position of the energy window on the
photopeak.
(viii) Defects in the collimator (extrinsic uniformity).
(ix) Radioactive contamination on the collimator or detector crystal.
(x) Crystal hydration.
(xi) Broken detector crystal (due to impact or thermal changes).
(xii) Improper QC procedure, including errors due to phantom preparation,
e.g. size of a point source, flood source filling, source positioning.
A test of flood field uniformity is the basic most sensitive QC test for the
gamma camera. This QC test should be considered as a first troubleshooting
QC test when an image artefact is encountered or suspected (seeexample
in Fig.I.8). However, if there is suspicion of a local artefact, relating to a
possible PMT malfunction, in the patients images, it may be the simplest
to first repeat an image after moving the patient within the field of view
(FOV); if the artefact moves with the patient, the problem is patient
oriented; if the artefact remains in the same location, then it is instrument
oriented. Further investigations can then be performed (e.g. by making a
uniformity QC test). An example of the effect of a defective PMT artefact
in static bone scans is shown in Fig.I.9.
696
FIG.I.8. (a)A clinical 111In somatostatin receptor study obtained with a single detector gamma
camera. The upper and lower abdomen in the anterior view (top two images), and the upper
and lower abdomen in the posterior view (bottom two images). Each image of the clinical
study showed two large, diffuse, circular colder areas (indicated by arrows). (b)Uniformity
image obtained after the clinical study, which shows two large cold areas with a hot border,
each due to a defective photomultiplier tube. The non-uniformities in this example were large
enough to be recognized in the patients images at the time of imaging. The images could,
therefore, be repeated on another gamma camera system. The problem occurred intermittently,
but the fault was never localized. The camera was finally replaced. (Example 2.2.8.6 in the
IAEA Quality Control Atlas for Scintillation Camera Systems.)
The appearance of an artefact in the uniformity QC test is dependent

on the problem. If the non-uniformity is diffuse or unclear, a sensitive
troubleshooting method is to make two further uniformity QC images
with asymmetrically positioned energy windows: with the energy window
set asymmetrically over the lower half of the photopeak and with the
energy window set asymmetrically over the upper half of the photopeak.
Non-uniformities are highlighted in such asymmetric images, with cold
areas in the one image corresponding to hot areas in the other image.
Asymmetric images highlight, for example, poor tuning, problems with an
energy correction map, ADC (analogue to digital converter) problems and
crystal hydration. FigureI.10 shows an example of the early appearance
of extensive non-uniformity patterns in images made with asymmetrical
energy windows. These artefacts were attributed by the service engineer
to separation of the light pipe from the crystal, which could not be rectified
by service and implied replacement of the whole detector. Six months later,
these artefacts became evident in the clinically used symmetrical energy
697
APPENDIX I
FIG.I.9. Static images of the skeleton after administration of 99mTc phosphonate. Images
obtained on camera A (top left posterior, top right right anterior, bottom left right
anterior oblique) show an area of apparent decreased activity in the lower spine that is
especially evident in the posterior and right anterior oblique views. As the cold area in the
lower spine was unusual, it was not considered to indicate pathology but to be an artefact.
The posterior skeleton was, therefore, imaged on camera B, and these images show a normal
99m
Tc-phosphonate distribution in the lower spinal column. Subsequently, a uniformity
image was obtained on camera A that demonstrated a defective photomultiplier tube that
corresponded to the area of decreased activity in the skeleton images. Camera A required
servicing before further clinical images were performed. Note: The study was reviewed before
the patient left the department. If a second camera had not been available, the patient could
have been shifted so as to image the lower skeleton in another part of the camera field of view.
(Example 2.2.8.5 in the IAEA Quality Control Atlas for Scintillation Camera Systems.)
window. Early observance of such a situation can assist with initiating

a replacement plan. An unusual and unexpected discovery of crystal
hydration in a new camera 3 months after installation is shown in Fig.I.11.
In this situation, detector replacement was required but this was within
the guarantee period. A dramatic example of hydration and poor tuning is
shown in Fig.I.12.
(d) Poor image spatial resolution and image contrast: Poor planar image spatial
resolution can be caused by:
(i) Too large a distance between the patient and collimator.
(ii) Poor linearity corrections of the detector: As visual evaluation of
linearity is subjective and difficult to assess, linearity should be
698
Counts
Counts
Energy (keV)
Energy (keV)
FIG.I.10. Top row: A series of routine intrinsic uniformity images obtained with
99m
Tc, uniformity correction turned off, with the energy window set symmetrically (left),
asymmetrically low (middle) and asymmetrically high (right) over the photopeak. The bottom
image is a repeat intrinsic uniformity image obtained 6 months later on the same gamma
camera. Each image was made with 5 Mcounts in a 256 256 matrix. The asymmetrical
low image shows a large diffuse rectangular hotter central area that corresponds on the
asymmetrical high image to a colder central area. Six months later, the same central colder
area is now visible on the image obtained with the symmetrical photopeak. This artefact was
caused by a separation of the light pipe from the crystal. This problem could not be fixed and
the whole detector required replacement. In this particular case, the detector replacement was
covered in the service contract. It would otherwise have been a very expensive repair. Note: The
colour scale used in these images is reversed between the first images (0100% counts=black
to white) and the image made 6 months later (0100% counts=white to black). Recording the
colour scale together with the images is essential, not only for quality control images but also
for clinical images. The colour scale and any colour enhancement should always be taken into
consideration when reviewing images.
quantified if software is available. FigureI.13 shows the results of

a 6-monthly QC test of spatial resolution and linearity using a slit
phantom, where the quantified National Electrical Manufacturers
Association linearity values were outside of the specifications in
both the X and Y directions, indicating that a new linearity map was
necessary.
(iii) Poor multiple energy window registration.
A decrease in image contrast in acquired images may be caused by:
(i) Incorrect position of the energy window and may be an operator error:
For example, this can occur if a test or calibration has been performed
699
Symmetric window: 140 keV
Tc-99m, 5 Mcounts, 256 x 256 matrix
Asymmetric window: 125 keV
Counts
APPENDIX I
Energy (keV)
FIG.I.11. Intrinsic uniformity images obtained at 3 months after installation of a new gamma
camera. The images were obtained with the uniformity correction turned off (but linearity
and energy corrections activated), using a 99mTc point source, 256256 matrix, 5 Mcounts
total. The left image was obtained with the 99mTc energy window set symmetrically over the
photopeak. It shows a suspicious small cold spot on the lower border. In order to investigate
this further, the energy window was offset on the lower half of the photopeak (seediagram).
The image obtained with this window setting (right) shows a distinct hot spot at the same
location as the cold spot on the left images, as well as two other small hot spots close by. This
is the result of crystal hydration. The detector can still be used at this moment in time, because
the hydrated areas are at the edge of the field of view. However, hydration will continue to
develop. The detector required replacement. In this case, the problem was discovered soon
after installation within the guarantee period, so that replacement could be made under the
guarantee. Note: If this situation is observed in an older gamma camera, a replacement
strategy for the detector must be planned. The development of hydration requires close
monitoring by weekly or monthly asymmetric uniformity images until replacement takes place.
with a 57Co source, and inadvertently the energy window not been
reset to 99mTc.
(ii) Performing an automatic peaking procedure with the patient as the
radioactive source: Owing to the large additional scatter component
in the photopeak, the window will automatically adjust too low over
the photopeak. The clinical image will, thereby, include unnecessary
scatter in the image.
In SPECT, a decrease in resolution and contrast may be related to:
(i) The imaging technique (e.g. excessively large radius of rotation, poor
choice of acquisition and reconstruction parameters).
700
Symmetric window: ~125 keV
Asymmetric window: ~156 keV
Counts
Energy (keV)
FIG.I.12. Periodic intrinsic uniformity images obtained with 99mTc, the uniformity correction
turned off, and asymmetric energy windows set low (left) and high (right) over the photopeak
(each image: 5 Mcounts, 256256 matrix). The images demonstrate extreme crystal hydration
over the whole field of view: small hot spots in the low asymmetric window correspond to small
cold spots in the high asymmetric window. The asymmetric images also show some poor tuning
(especially in the top right corner). The extent of the hydration indicates that this detector
requires replacement.
(ii) Inadequate instrument calibration (offset in centre of rotation (in X or

Y directions), detector tilt in Y direction, poor alignment of multiple
detector heads, inadequate uniformity correction).
(iii) Artefacts in acquired data (missing projections, PMT artefact).
(iv) Artefacts in reconstructed data (e.g. ring artefacts from
non-uniformity).
Troubleshooting artefacts observed in SPECT images may simply involve
a test SPECT acquisition of a point source placed off-axis (seeFig.I.7)
(e.g. for assessing problems of motion between detector heads), or might
be more intensive involving a test SPECT acquisition of a cylindrical bottle
or a SPECT phantom (e.g. assessing ring artefacts observed in clinical
images). The recalibration of the uniformity correction map or the centre
of rotation and head alignment (and head alignment in multiple detector
systems) may solve the problem. However, a follow-up QC test following
recalibration is always required, in order to check that the problem has been
solved. FigureI.14 is an example where a recalibration of head alignment
was insufficient, and a remaining problem of detector head tilt required
a service visit. Further awareness is still required if the problem is an
intermittent fault and not solved by recalibration alone (as in the situation
of the example in Fig.I.7).
701
APPENDIX I
FIG.I.13. Routine 6-monthly quality control test of intrinsic linearity and spatial resolution
of a small field of view gamma camera, using a slit phantom with 1cm spacing between slits,
99m
Tc point source, 1024 1024 matrix (pixel size: 0.29 mm). The acquired images were
quantified within the indicated rectangles. The spatial resolution was within specification.
However, linearity (absolute deviation: Abs Dev; maximum line deviation: Max LineDev)
was out of specifications in both the X and Y directions. The linearity correction maps
needed recalibration. NEMA: National Electrical Manufacturers Association; FWHM: full
width at half maximum; FWTM: full width at tenth maximum; UFOV: useful field of view;
CFOV: central field of view.
(e) Clinical investigations: A review of data acquired is essential before

processing and quantification:
(i) For SPECT data, a cine, sinogram and linogram can suffice to review
the clinical data for patient movement, missing projections, instability
(an artefact that appears in only some projections) and inadequate
continuity of data from multiple detectors. If the review reveals such
errors, then the study may need to be repeated.
Note that an artefact due to patient movement will be imaged by all
detectors at the same time, so that the movement artefact will repeat. If
a movement artefact appears only in one detector of a multiple detector
system, then the problem is probably due to the instrument (as an example,
see Fig.I.7).
702
FIG.I.14. Top: Quality control images of a SPECT myocardial perfusion study using a dual
head gamma camera in a 90 configuration, 180 rotation (90 per head), 128128 matrix,
4.8mm pixel size. Images: left one projection of the acquired data, middle sinogram (X)
of a profile over the myocardium (shown in the left image), right linogram (Y) for the same
profile. There is a discontinuity between detector 1 and 2 visible on both the sinogram and
linogram. Bottom: After recalibration of the centre of rotation and head alignment (first
trouble-shooting technique applied), a test acquisition was made of a 99mTc point source placed
off-axis (left image projection). The sinogram (middle) shows no discontinuity, whereas the
linogram (right) shows both discontinuity and slope in the data particularly evident in the
quantified offset graph. The problem was a detector head tilt in both heads, which required a
service.
I.3. MINIMIZING PROBLEMS

Problems can occur at any time, but taking appropriate precautions
canminimize the likelihood.
I.3.1. Siting and room preparation
The location of an instrument and good preparation prior to installation are
vital first steps. Particular care is needed to choose a location suitable for the
703
APPENDIX I
specific instrument in order to avoid interference from X rays, magnetic fields

(magnetic resonance imaging), radiotherapy machines, radioactive sources
from the radiopharmacy, injection room or radioactive patients (such as from
radionuclide therapy or positron emission tomography (PET)).
Considerations for room preparation should include the following:
(a) Necessary wall shielding from extraneous radiation and magnetic fields.
(b) Floor weight support and floor levelling.
(c) Continuous stable electrical power supply: Consider connecting the
instrument to an emergency power supply and installing an uninterruptible
power supply (UPS) (seeSections7.6.2 and I.3.2). Consider the electrical
conditioning, grounding and safety.
(d) Sufficient strategically placed power outlets for peripherals.
(e) Lighting and switches (to exclude electrical interference with equipment).
(f) Window placement with respect to the instrument position to avoid drafts
and influence from direct sunlight. (Particular care is needed for the gamma
camera with respect to exposing the crystal to a sudden temperature
change such as might happen during collimator change and intrinsic QC
measurements.)
(g) Stable air conditioning with respect to temperature (maximum,minimum,
fluctuating temperatures) and humidity (non-condensing): Consideration
should be given to these aspects not only during working hours, but also
outside of working hours, including at weekends. A major hazard to a
gamma camera crystal is a rapid change of temperature: a rule of thumb is
that the temperature should not change more than 4C over 1 h.
(h) Dust free environment: It is generally not possible to achieve a dust free
environment in a hospital. However, maximizing a dust free environment
should be aimed for, especially for the computers and picture archiving and
communication system.
(i) Positioning of the instrument within the room tominimize interference
from external radioactive sources.
I.3.2. Electrical power conditioning
The stability and correct voltage level of the electrical supply is crucial
to reducing the likelihood of obtaining an instrument malfunction. This may be
achieved by use of a surge protector, a constant voltage transformer or a UPS
(battery backup), the choice being dependent on the type of equipment and local
environmental requirements.
The UPS is essential where power failures or major power dips and surges
occur, in order to avoid the disastrous failure of instruments and computers and
704
the breakdown of components. Even if the instruments regular power supply

is connected to an emergency power supply, the interval between failure of the
regular power supply and the initiation of the emergency power supply may
produce a power dip sufficient for the instrument to halt or a circuit breaker to
switch off (particularly disruptive when occurring during a patient study; see also
Fig.I.6).
The investment for a UPS must be considered when requisitioning
and purchasing the instrument. A UPS should be connected to all sensitive
instruments that are required for daily routine patient care. The UPS specification
is dependent on the instruments power requirement and the local power situation.
The UPS may be needed to ensure that, in the event of a power failure during
working hours, the instrument can be manually shut down in the correct way. The
UPS may be needed only to bridge the gap between a regular power failure and
the switch to the emergency power supply.
The electrical conditioning includes appropriate grounding, and shielding
of cables, especially for signal cables and data transmission cables.
Electrostatic disturbances can beminimized by adequate humidity control
(air conditioners), and antistatic work surfaces and floor covering.
I.3.3. Regular preventive maintenance
Regular preventive maintenance and a service contract can help not only
tominimize the chance of an unexpected problem occurring, but also tominimize
the down time when a problem has occurred. The expense of a service contract
can be considered as an insurance policy. A service contract should ensure fast
response and priority access to spare parts. The service may include remote
computer login and access. In-house access to trained personnel is also essential
for first-line troubleshooting of electrical failures, computer and network failures,
and mechanical failures. Without any access to appropriate support, a problem
can take a considerable time to be solved, add extra expenses, and become a
major obstacle to high quality and continuous patient care.
I.3.4. Acceptance testing and routine quality control testing
Thorough and careful acceptance testing is the first step towards ensuring
that an instrument is performing according to specifications and as expected
for clinical use. Any problems or suspected problems encountered at this early
stage require instant rectification as the instrument is still under the guarantee
period (seethe example of crystal hydration observed at 3 months in Fig.I.11).
Replacement of any defective component must be initiated. The collimator is
particularly sensitive to damage from transport and must be tested carefully at
705
APPENDIX I
Physical set-up:
Collimator
200 MBq Tc point source
placed at 4 m from collimator off centered
1024 1024 matrix
10 Mcounts
99m
Point source
Point spread image
34 m
FIG.I.15. Acceptance testing of a medium energy collimator using a distant point source of
99m
Tc (acquisition parameters given above). The point source was positioned first to image
the right part (left image) and then the left part of the collimator (right image). There are
vertical discontinuities evident, probably from the manufacturing process. This collimator was
replaced within the guarantee period. Note: This is a sensitive method for testing a collimator
for hole angulation problems and for any suspected damage. A large distance between the
source and collimator is essential. This test supplements a high count system uniformity test.
acceptance and at any time when damage is observed or suspected. An example

of a defective collimator discovered at acceptance testing is shown in Fig.I.15.
Routine QC tests are performed on the gamma camera and SPECT system
in order to assess performance of the instrument at a specific moment in time.
They are intended to reassure the user that performance up to that moment is
satisfactory. Monitoring the results of successive QC tests can indicate a stable
functioning condition, deterioration or an impending problem. A database of
results is recommended. The visual as well as quantitative results must always
be reviewed together. FigureI.16 illustrates a situation in which the quantitative
uniformity value appears to be acceptable, whereas the image shows there is
a PMT problem. Routine QC tests are valuable in troubleshooting, and should
neither be underestimated nor neglected. The results of QC tests can highlight the
underlying problems.
706
FIG.I.16. Routine intrinsic uniformity image with quantification. The image was obtained
with a 99mTc point source, symmetrical energy window over 140keV, 30 Mcounts. The image
shows a hot semicircular area in the lower right field of view. The quantification indicates
that the uniformity is acceptable. However, not indicated in the results, the uniformity
calculation refers only to the central field of view. The artefact was due to a malfunctioning
photomultiplier tube. Note: This camera required a service but could continue to be used with
caution because of the lateral location of the defect. This example illustrates that it is essential
to review together both image and quantification, and to understand the parameters provided
in the results.
I.4. IMAGE ARTEFACTS IN PET/CT

PET and combined PET/computed tomography (CT) require users to
develop skills in recognizing a range of artefacts which are quite distinct
from those which may be seen in SPECT or SPECT/CT. PET and SPECT
reconstruction do have aspects in common, resulting in analogous methods of
recognition; however, the artefacts themselves may appear quite different due to
intrinsically different modes of acquisition. PET scanners usually employ a fixed
full-ring detection system, unlike SPECT which has a rotating gamma camera,
thus eliminating the need for a centre of rotation correction and its associated
artefacts. In a typical PET system, a ring of detectors surrounds the patient, each
of which simultaneously and independently acquires data. In addition, there is
no collimator used in 3D PET, leading to a vast increase in scanner sensitivity
such that acquisition times are generally shorter and whole body scans are the
norm. The use of very many individual detectors in PET implies that cameras
withminor defects can be tolerated unlike in SPECT, where a defect has a
variable impact depending on its location (greater impact towards the centre of
the FOV) but may still be usable if the defect is towards the edge of the FOV.
707
APPENDIX I
PET is most often performed with an accompanying CT scan, usually acquired

using a hybrid scanner where the CT component can be used diagnostically.
This requires bed translation between the PET and CT scanners, whereas some
SPECT/CT scanners use an integrated low-end CT scanner that is co-located
with the gamma camera detectors within the gantry and does not require any bed
translation.
The medical physicist needs to be able to derive what the underlying
problem is from the artefact, whether it is of a hardware or software nature. This
task can be difficult owing to the very diverse way in which scanner problems
can present themselves as artefacts. It is useful to classify PET/CT artefacts
into the following categories: tomographic artefacts, attenuation correction
(AC) artefacts, co-registration artefacts and movement artefacts. Explanations
and examples of each are given below, and IAEA Human Health Series No. 27
provides further information and examples.
I.4.1. Tomographic artefacts
Tomographic artefacts are those which appear when some fundamental
aspect of the tomographic system performs below specification or else fails
entirely. Problems in the tomograph lead to systematic image abnormalities
that occur regardless of the type of acquisition being performed. One such
abnormality is due to incorrect normalization. FigureI.17 demonstrates an
artefact that was created when the normalization of the tomograph had been
corrupted. Normalization corrects for the sensitivity difference between
different lines of response (LORs). Sensitivity differences are caused by both
a geometric distortion, which needs to be measured only once at the factory,
and by detector efficiency variations that can change with time and must be
periodically recalibrated. Normalization errors occur in the projection space and
appear as circular defects in the transaxial reconstructed space. In the example of
Fig.I.17(a), the artefact is seen to be repeated in each bed position of the whole
body acquisition, indicating that there was a problem with the tomograph itself.
The daily quality assurance routine is a good way to detect unexpected
sudden normalization errors. Some quality assurance routines involve scanning
a cylindrical phantom filled uniformly with radioactivity (e.g. 68Ge, 18F). Such a
cylindrical phantom is designed to be large enough to cover many of the potential
LORs in the system. The artefacts due to normalization errors seen in the clinical
images of Fig.I.17(a)were clearly visible in the uniformity QC image, as shown
in Fig.I.17(b).
Geometry is often the key in diagnosing tomographic artefacts, as can be
seen in Fig.I.18, where there has been failure of a detector block leading to a
distinctive pattern in the sinogram of the QC uniformity image. Detector block
708
FIG.I.17. (a) Clinical whole body images obtained on a PET system in which the
normalization correction was corrupted, but not known at the time. The sagittal view shows
a pattern of repetitive cold horizontal stripes at consistent locations within each of the bed
positions. The periodic nature of the artefact is a sign that the problem is associated with
the system rather than this particular patient or acquisition. (b)Sagittal view of a uniformity
quality control check of the PET system acquired using a uniformly filled cylindrical phantom.
The image shows cold stripes indicative of errors in the normalization table. This quality
control image was obtained after the clinical images revealed the artefacts shown in (a).
The quality control image shows several cold streaks which indicate that the problem is most
likely a corrupted normalization file. Normalization was recalibrated before further patient
acquisitions were performed. (Courtesy of the Department of Nuclear Medicine, Monte Tabor
So Rafael, Brazil.)
failure may not contraindicate the clinical use of a PET system since modern
scanners have many detectors and the absence of one block may have little
statistical impact. Examination of the sinogram (also in clinical images) is a good
way to test for block failure, as it appears as a distinctive diagonal streak on the
sinogram.
I.4.2. Attenuation correction artefacts
AC artefacts occur when the CT AC algorithm leads to a hot or cold spot
in the attenuation corrected reconstructed data. AC effectively increases the
counts in each voxel in proportion to the total attenuation along all LORs that
pass through that voxel. When the CT image shows a highly attenuating material
in a group of voxels, then the total counts along all lines of response that pass
through those voxels are increased, and the group of voxels appears hot. This is
709
APPENDIX I
FIG.I.18. Sinogram from a PET system that has a detector block failure. The diagonal streak
that is clearly visible is the pattern created when one detector block has failed and causes
many lines of response to be zero. The failed detector block creates several blank lines of
response at every projection angle at incremental radial positions and the result is a diagonal
streak. With only one streak visible, and the fact that the streak is several pixels wide, it would
be appropriate to assume that a whole detector block has failed. The noticeable width in
the streak occurs because each detector block contains many individual detector elements.
Multiple simultaneous detector block failure is unlikely in a system which has regular quality
assurance tests. This system is still acceptable for clinical use because there are many detector
blocks in a PET system and the loss of one block results in a drop in sensitivity of only ~0.5%.
particularly noticeable where the patient has metal implants or has taken contrast
media. The attenuation of metal and contrast at the CT energy does not relate
linearly to the attenuation at annihilation photon energy. In this situation, the AC
is overestimated and a hot spot appears erroneously at the point where the metal
or contrast media is found. FigureI.19 demonstrates a contrast artefact leading
to a hot spot that appears cold on the corresponding non-AC PET image. The
non-AC images are often a key component in recognizing metal or contrast based
artefacts, but the presence of streaks in the CT image is also a warning sign.
The non-AC images should always be reviewed whenever any dubious finding is
suspected in the AC images.
Another AC artefact is due to truncation, where the CT and PET FOVs
are not the same size, so that parts of the anatomy outside the CT FOV are not
corrected for by the AC algorithm. This often occurs when the patients arms
(which are raised above the head during the acquisition) are outside the CT FOV
and a cold stripe appears across the patients head in the AC images. In Fig.I.20,
a cold stripe is prominent in the AC images but not visible in the non-AC images.
Some PET/CT systems include software that can reconstruct the truncated CT
data to increase the FOV and, thereby, reduce the severity of the artefact.
710
FIG.I.19. (a)CT attenuation corrected image of a patient showing a focal hot spot (indicated
by the arrow). (b)Non-attenuation corrected image. The hot spot is no longer visible. (c)CT
image showing a high density artefact from barium contrast pooled in the bowel. The artefact
appears to be a region of high attenuation and the reconstruction algorithm overcompensates
and creates a false hot spot. On the non-attenuation corrected image, the hot spot is entirely
gone. These high density material artefacts are very common and the user should always
examine the non-attenuation corrected image to check for the presence of such artefacts. Clues
can also be found by examination of the corresponding location on the CT. (Courtesy of the
Department of Nuclear Medicine, Memorial Sloane Kettering Cancer Center, New York.)
I.4.3. Co-registration and motion artefacts

Problems in co-registration in PET/CT are common and can be due to a
system error or caused by movement of the patient. The system must be tested
and recalibrated periodically, whereby a transform matrix is created to co-register
the PET and CT data. Small errors in co-registration can often be seen in the head
where the brain does not correctly fit inside the skull. Errors in the co-registration
can occur either suddenly or gradually and can be a sign that there is a problem
with the mechanism that controls the bed motion. Regular QC is required.
Alignment errors originating from patient motion are problematic and
can have an effect on the medical interpretation of the image. The effect of an
alignment error is demonstrated in Fig.I.21, where the patients head has moved
during data acquisition causing a misalignment between the PET and CT data.
In the attenuation corrected images, the cortical uptake appears asymmetrical,
but it can be seen from the fused PET and CT image that this was caused by a
mis-registration error.
711
APPENDIX I
FIG.I.20. Clinical [18F]-fluorodeoxyglucose images of the head and thorax from

a PET/CT system. Top row: non-attenuation and scatter corrected PET images;
middle row: the corresponding slices with attenuation and scatter correction; bottom row:
CT images. The patients arms have been truncated and extend beyond the CT field of view
(arrow, bottom row). Although the truncation was relatively localized, using these CT data
to correct for attenuation and scatter produced more extensive errors. This is shown by
the cold band in corrected coronal and sagittal images of the head (arrows, middle row).
The extent of this artefact may be due to an error in scaling of the scatter correction. The
non-attenuation corrected images do not show the cold band. This example also demonstrates
the essential value of comparing images with and without attenuation correction. (Courtesy
of R. Boellaard, Department of Nuclear Medicine and PET Research, VU University Medical
Centre, Amsterdam, Netherlands.)
Patient motion artefacts can be easily missed and lead to an incorrect

diagnosis. In whole body PET scans, it is possible for the patient to move during
the acquisition so that some part of the anatomy is accurately registered between
PET and CT, while in another part of the anatomy the registration is poor.
When not noticed by the operator and the reporting doctor, these artefacts can
be misinterpreted as pathological uptake or be mistaken for a problem with the
system. An example of this is shown in Fig.I.22, where the patient moved towards
the end of the scan, causing an erroneous hot spot to appear in the wrong place.
Another movement artefact often seen is due to respiratory motion. PET
images are acquired over many respiratory cycles, such that the final image is
an average activity distribution across the respiratory cycle. The CT images are
712
FIG.I.21. Apparent asymmetrical uptake of 18F-fluorodeoxyglucose in the brain (left) caused

by the slight misalignment of the PET and CT images (right). The non-attenuation corrected
image did not show this asymmetry. These images form part of a whole body acquisition, which
commenced at the thigh and moved up towards the head, which was the last part of the scan.
Movement in the lower part of the body was not evident.
acquired far more quickly and, thus, demonstrate blurring over only a small
component of the respiratory cycle. This can create a mis-registration and
blurring in the AC PET images at the boundary of lung and liver. This kind
of artefact can have significant clinical implications when a tumour is found
near the border between the lungs and the liver. FigureI.23 shows an example
where a lesion in the liver is incorrectly located in the lungs. This is due to the
CT being acquired during full inspiration (the patient has taken a deep breath,
pushing the diaphragm down and displacing the liver caudally), as opposed to
the PET which is time averaged over regular tidal breathing, resulting in a severe
mis-registration between the functional and anatomical location of the lesion.
Conversely, Fig.I.24 shows an example where a tumour is incorrectly located in
the liver due to a respiratory motion artefact.
Respiratory motion artefacts can also be seen in the CT image itself where
the liver is not correctly rendered during reconstruction of the CT data because
the patient is breathing during acquisition. This can be seen in Fig.I.25 where
there is a characteristic artefact repeated along the axis of motion, leading to
unclear definition of organ boundaries.
I.5. IMPORTANCE OF REGULAR QUALITY CONTROL
Regular QC procedures vary between scanner vendors; however, daily
QC often requires checking the gantry status (voltages, temperatures, etc.) and
713
APPENDIX I
(a)
(b)
(c)
(d)
FIG.I.22. (a)A whole body PET/CT scan shows a point of focal uptake in the upper torso.
(b)Separate head and neck acquisition of the same patient the focal hot spot seems to have
moved to a different location. (c)Whole body fused images show good registration between
PET and CT in the bladder, spine and heart, and so it was assumed that the images were
correctly registered; however, closer inspection of the head shows a clear mis-registration
(indicated by the red arrow). (d)Fusion between PET and CT in the separate head and neck
view shows good registration. The operator did not closely inspect the head and neck portion
of the whole body view since there was a separate head and neck acquisition; however, the
patient moved during the scan, probably by rotating the head. Had there not been a separate
head and neck view, the doctor would have reported the focal uptake as metastatic cancer.
This image was reported to the staff physicist as a problem with the system; however, it was in
fact operator related. (Courtesy of the Nuclear Medicine Department, St. Vincents Hospital,
Darlinghurst, Australia.)
generation of the normalization from a high count emission image to ensure

image quality (including checking a number of parameters, such as block noise
and efficiency, scatter ratio, time alignment, etc.). The system vendor should
provide a daily QC package that automates all of the above requirements so
that QC can be performed quickly by the operator before the commencement
of scanning. The daily QC procedure should produce a report indicating any
unsatisfactory results which require further attention and allow for systematic
monitoring of the scanner.
The characteristics of PET/CT systems allow for quantitative imaging that
can display the absolute concentration of tracer in the subject. The ramification
714
FIG.I.23. (a) Transaxial PET and CT images show a focal lesion, which appears to be in
the lung. (b)The CT image shows the same lesion appearing to be in the liver. (c)Coronal
view of PET and fused PET/CT where the lesion appears largely displaced from the liver. This
problem occurs because the CT acquisition is very brief and captures the liver at one point
in the respiratory cycle (in this case, full inspiration such that the diaphragm has displaced
the liver caudally), while the PET acquisition is much longer and averaged over the whole
respiratory cycle. The activity of the lesion is underestimated due to an attenuation correction
artefact. This artefact stems from the fact that lung tissue is less attenuating than liver tissue
and so the reconstruction process under-corrects for the attenuation of the signal from the
lesion when it is assumed to be in the lung. The lesion demonstrates intense uptake and so is
still highly visible despite the attenuation correction artefact.
715
APPENDIX I
FIG.I.24. (a)Coronal PET image showing an area of focal uptake that appears to be both in
the liver and in the lung, as well as another larger area that appears to be entirely in the lung.
(b)Fused coronal PET/CT image showing a mis-registration in the larger lesion. Both of these
lesions are entirely contained in the lung and the elongated appearance of the smaller lesions
is an artefact created by respiratory motion.
of this for the physicist is the necessity to regularly perform a check of the
standardized uptake value (SUV). The SUV is a quantitative parameter often
quoted in clinical PET reporting that represents the uptake of activity in a lesion
relative to background or healthy tissue (which should have an SUV = 1). The
SUV measure is highly dependent on patient preparation, scanning protocol and
reconstruction technique, and should be used with caution. It also relies heavily
on accurate scanner calibration relative to the department dose calibrator, which
allows PET data to be quantitative. Despite these difficulties, this index is often
used by physicians for indicating abnormal uptake and, in particular, monitoring
patient response during treatment by comparison of the SUV at baseline to
the SUV during or after therapy. As such, the physicist must verify that such
measures are accurate. A monthly check of the scanner SUV should be performed
using a phantom of known volume, which, if activity is homogeneous and the
scanner and dose calibrator are correctly calibrated, should produce an SUV = 1.
Erroneously low SUVs may indicate that the physicist needs to recalibrate the
dose calibrator and PET scanner, through the calculation of a new calibration
factor.
716
FIG.I.25. (a)A sagittal image shows a step-like artefact in the liver. (b)The same step-like
artefact is seen on the coronal views. This problem is caused by the patient breathing during
the CT acquisition and distorting the size of the liver. These artefacts are very common and can
be compensated for by using a breath-hold technique.
This discussion of artefacts in PET/CT is by no means exhaustive and the

reader is referred to the IAEAs PET/CT Atlas on Quality Control and Image
Artefacts for a more comprehensive set of examples. As with all other instruments
of the department, developing expertise in PET and PET/CT is essential for
trouble-shooting and recognizing artefacts. Regular QC is a crucial factor
in reducing artefacts due to the tomographic and/or CT system. Users of PET
and PET/CT should be alert at all times to unexpected artefacts. A comparison
between attenuation and scatter corrected images with non-corrected images
should be part of routine clinical practice.
717
APPENDIX I
BIBLIOGRAPHY
BUSEMANN SOKOLE, E., PACHCNSKA, A., BRITTEN, A., Acceptance testing for
nuclear medicine instrumentation, Eur. J. Nucl. Med. Mol. Imaging 37 (2010) 672681.
BUSEMANN SOKOLE, E., et al., Routine quality control recommendations for nuclear
medicine instrumentation, Eur. J. Nucl. Med. Mol. Imaging 37 (2010) 662671.
INTERNATIONAL ATOMIC ENERGY AGENCY (Vienna)
Handbook on Care, Handling and Protection of Nuclear Medicine Instruments (1997).
IAEA Quality Control Atlas for Scintillation Camera Systems (2003).
Quality Assurance for SPECT Systems, IAEA Human Health Series No. 6 (2009).
Quality Assurance for PET and PET/CT Systems, IAEA Human Health Series No. 1 (2009).
PET/CT Atlas on Quality Control and Image Artefacts, IAEA Human Health Series No. 27
(2014).
718
Appendix II
RADIONUCLIDES OF INTEREST IN DIAGNOSTIC
AND THERAPEUTIC NUCLEAR MEDICINE
Energy of
emissiona
(MeV)
Abundance
(in same order
as emissions)
Radionuclide
Half-life
Principal
radiations
emitted
11
20.5min
+ ()
0.39
100%
13
9.97min
+ ()
0.49
100%
15
122.2 s
+ ()
0.74
100%
18
109.8min
+ ()
0.24
96.9%
32
14.2 d
0.695
100%
51
27.7 d
0.32
9%
57
271.7 d
0.122
86%
62
9.7min
64
12.8 h
(),
0.28 ( )
0.19 ()
17% (+)
39% ()
67
62 h
0.91 (1)
0.93 (2)
0.184 (3)
0.121 (1)
0.154 (2)
0.189 (3)
7% (1)
16% (2)
49% (3)
56% (1)
23% (2)
20% (3)
67
78 h
0.093
0.184
0.300
0.394
38%
24%
16%
4%
68
68min
+ ()
0.74
88%
81m
13.1 s
0.191
100%
82
75 s
+ ()
1.4
96%
0.585
100%
0.897 (max.)
0.909
22.3%
99%
C
N
O
F
P
Cr
Co
Cu
Cu
Cu
Ga
Ga
Kr
Rb
89
50.5 d
89
78.4 h
+ ()
Sr
Zr
2.93 (max.)
97%
719
APPENDIX II
Energy of
emissiona
(MeV)
Abundance
(in same order
as emissions)
Radionuclide
Half-life
Principal
radiations
emitted
90
64 h
0.93
100%
99m
Tc
361.2min
0.1405
89%
111
In
67.4 h
0.172
0.247
89%
94%
123
13 h
0.159
97%
124
4.2 d
+ (),
0.97 (+1)
0.69 (+2)
0.603 (1)
1.69 (2)
11% (+1)
12% (+2)
62% (1)
0% (2)
125
60.2 d
0.036
7%
131
8.04 d
0.19 ()
0.364 (1)
0.637 (2)
90% ()
83% (1)
7% (2)
131
9.7 d
0.353
100%
I
I
I
I
Cs
133
5.25 d
0.10 ( 1)
0.081 (1)
100% (1)
37% (1)
153
1.95 d
0.20 (1)
0.23 (2)
0.27 (3)
0.070 (1)
0.103 (2)
32% (1)
50% (2)
18% (3)
5% (1)
28% (2)
166
26.8 h
0.70 (1)
0.65 (2)
50% (1)
49% (2)
169
9.4 d
0.10
100%
177
6.73 d
0.15 (1)
0.12 (2)
79% (1)
9% (2)
186
3.78 d
1.07 (1)
0.93 (2)
0.140 ()
74% (1)
21% (2)
9% ()
188
17.0 h
0.79 ()
0.73 (2)
70% (1)
26% (2)
198
2.7 d
0.32 ()
0.40 ()
99% ()
96% ()
Xe
Sm
Ho
Er
Lu
Re
Re
Au
720
RADIONUCLIDES OF INTEREST IN DIAGNOSTIC AND THERAPEUTIC NUCLEAR MEDICINE
Energy of
emissiona
(MeV)
Abundance
(in same order
as emissions)
Radionuclide
Half-life
Principal
radiations
emitted
201
Tl
73 h
, X
0.167 ()
0.070 (X1)
0.080 (X2)
8% ()
74% (X1)
20% (X2)
211
At
7.14 h
5.868
42%b
212
1.01 h
a,
6.1 ()
2.25 ()
34% ()
55% ()
213
45.6min
5.55.9 ()
100% ()
223
11.4 d
5.55.7 ()
0.154 ()
97% ()c
6% ()
Bi
Bi
Ra
Average energy of emission.

Astatine-211 undergoes a branched decay 42% corresponds to direct emission. The
other 58% decays via electron capture to 212Po, which also decays by emission. As such,
the net effect is one emission per decay.
c
Radium-223 has a complex decay chain, and when it comes to rest, it results in the
emission of four particles per decay.
b
721
ABBREVIATIONS
AAPM American Association of Physicists in Medicine
AC
attenuation correction; alternating current
ACR American College of Radiology
ADC
analogue to digital converter
ADT
admission, discharge, transfer
AFOV
axial field of view
ALARA
as low as reasonably achievable
ANSTO Australian Nuclear Science and Technology Organisation
APD
avalanche photodiode
ASCII American Standard Code for Information Interchange
BED
biological effective dose
BGO
bismuth germanate
-methyl-p-iodophenylpentadecanoic acid
BMIPP
BMP bitmap
BSS Basic Safety Standards
CDR
collimatordetector response
CDF
cumulative density function
CERN European Organization for Nuclear Research
CF
calibration factor
CFD
constant fraction discriminator
CFOV
central field of view
CIE International Commission on Illumination
CLUT
colour lookup table
CMOS
complementary metal oxide semiconductor
CMS
colour management system
CMYK
cyan, magenta, yellow, key (black)
COTS
commercial off the shelf
counts per minute
cpm
counts per second
cps
CR
contrast ratio
CRT
cathode ray tube
CT
computed tomography
CZT
cadmium zinc telluride
DC
DCT
DDL
direct current
discrete cosine transform
digital driving level
723
ABBREVIATIONS
DIB
device independent bitmap
DICOM Digital Imaging and Communications in Medicine
DMSA
dimercaptosuccinic acid
DNA
deoxyribonucleic acid
dpi
dots per inch
DPM
disintegrations perminute
DRL
diagnostic reference level
DSB
double strand break
DTPA
diethylenetriaminepentaacetic acid
DVH
dosevolume histogram
EANM European Association of Nuclear Medicine
EBRT
external beam radiotherapy
ECG electrocardiogram
EDV
end diastolic volume
EGF
epidermal growth factor
EGS
electron gamma shower
eh electronhole
EM
expectation maximization
ERPF
effective renal plasma flow
ESV
end systolic volume
FBP
filtered back projection
FDA Food and Drug Administration
FDG fluorodeoxyglucose
FFT
fast Fourier transform
FORE
final rebinning algorithm
FOV
field of view
FWHM
full width at half maximum
FWTM
full width at tenth maximum
GFR
GIF
GPU
GSDF
GSO
glomerular filtration rate

Graphics Interchange Format
graphical processing unit
grey scale standard display function
gadolinium oxyorthosilicate
HMPAO
hexamethlypropyleneamine oxime
HPMT
hybrid photomultiplier tube
HTML Hypertext Markup Language
HU Hounsfield unit
724
ABBREVIATIONS
HVL
half-value layer
ICC International Color Consortium

ICRP International Commission on Radiological Protection
ICRU International Commission on Radiation Units and Measurements
IEC International Electrotechnical Commission
IEEE Institute of Electrical and Electronics Engineers
IPS
in-plane switching
IQ
image quality
ISO International Organization for Standardization
JFIF
JND
JPEG
JPEG File Interchange Format

just noticeable difference
Joint Photographic Experts Group
LCD
liquid crystal display
LET
linear energy transfer
LOR
line of response
LQ linearquadratic
LR
luminance ratio
LS
least squares
LSF
line spread function
LSO
lutetium oxyorthosilicate
LUT
lookup table
LZW LempelZivWelch
macroaggregate of albumin
MAA
MAG3 mercaptoacetyltriglycine
maximum a posteriori
MAP
Monte Carlo N-particle transport code
MCNP
microchannel plate
MCP
mean free path
MFP
MIBG metaiodobenzylguanidine
MIBI methoxyisobutylisonitrile
MIP
maximum intensity projection
medical internal radiation dose
MIRD
MLEM
maximum-likelihood expectation-maximization
MR
magnetic resonance
magnetic resonance imaging
MRI
medical record number
MRN
mean square error
MSE
725
ABBREVIATIONS
MTF
MUGA
MWPC
modulation transfer function

multiple-gated acquisition
multiwire proportional chamber
NEA
NECR
NEMA
NET
NHEJ
NIST
NPL
NTCP
NURBS
negative electron affinity

noise equivalent count rate
National Electrical Manufacturers Association
neuroendocrine tumour
non-homologous end joining
National Institute of Standards and Technology
National Physical Laboratory
normal tissue complication probability
non-uniform rational B-spline
OER
OSEM
oxygen enhancement ratio

ordered-subsets expectation-maximization
PACS
picture archiving and communication system
para-amino hippurate
PAH
Pan American Health Organization
PAHO
profile connection space
PCS
Portable Document Format
PDF
perceived dynamic range
PDR
positron emission tomography
PET
photomultiplier tube
PMT
POPOP para-phenylene-phenyloxazole
point spread function
PSF
point source response function
PSRF
plasma volume
PV
PVE
partial volume effect
QA
QC
QMS
quality assurance
quality control
quality management system
RAMDAC
RAMLA
RBE
RC
RGB
RIS
random access memory digital to analogue converter

row-action maximum-likelihood algorithm
relative biological effectiveness
recovery coefficient
red, green, blue
radiology information system
726
ABBREVIATIONS
RIT radioimmunotherapy
ROI
region of interest
RPO
radiation protection officer
RPP
radiation protection programme
RSS Real Simple Syndication
SI International System of Units
SiPM
silicon photomultiplier tube
SNR
signal to noise ratio
SPECT
single photon emission computed tomography
SPR
scatter to primary ratio
SSB
single strand break
SUV
standardized uptake value
TCP
TDC
TEW
TIFF
TOF
TLG
TVL
tumour conrol probability

time to digital converter
triple energy window
Tagged Image File Format
time of flight
total lesion glycolysis
tenth-value layer
UFOV
useful field of view
UNSCEAR United Nations Scientific Committee on the Effects of Atomic
Radiation
UPS
uninterruptable power supply
UTF Unicode Transformation Format
World Health Organization
WHO
weighted least squares
WLS
WYSIWYG what you see is what you get
XML Extensible Markup Language
727
SYMBOLS
Roman symbols
year (unit of time)
acceleration; area; specific activity
a
time integrated activity coefficient
A
ampere (SI unit of current)
atomic mass number
A

ngstrm (unit of distance: 1 = 1010 m)

A
cumulated activity
A activity
a
a
b
Bq
barn (unit of cross-section)

becquerel (SI unit of activity)
c
C
C
C
cd
Ci
speed of light
capacity; coulomb (SI unit of charge)
degree Celsius (unit of temperature)
activity concentration; counts
luminous intensity: candela
curie (unit of activity: 1 Ci = 3.7 1010 Bq)
d
d
D
day (unit of time)

depth; distance
dose; thickness

D
dose rate
dB decibel
e
electron charge
E
effective dose; electric field; energy
Eab
absorbed energy
EB
binding energy
EK
kinetic energy
Etr
transferred energy
Eg
photopeak energy
E()
committed effective dose
E ab
mean absorbed energy
E tr
mean transferred energy
eV electronvolt
729
SYMBOLS
F
farad (SI unit of capacitance)
F Fano factor; force
G
Gy
gravitational constant
gray (SI unit of dose)
h
h
H p
HT
HT()
Hz
hour (unit of time)

Plancks constant
personal dose equivalent
equivalent dose
committed equivalent dose
unit of frequency
electric current; intensity
j
J
current density
joule (SI unit of energy)
K
kelvin (SI unit of thermodynamic temperature)
K kerma
kg
kilogram (SI unit of mass)
l length
L
litre (unit of volume)
L
luminance
m
metre (SI unit of length)
m mass
ma
atomic mass in atomic mass units u
me
electron rest mass; positron rest mass
m n
neutron rest mass
m p
proton rest mass
M
nuclear mass
mol
amount of substance: mole
N
newton (SI unit of force)
N
number of counts; number of neutrons in an atom
Na
number of atoms
NA Avogadros number
Nel
number of electrons
Nph
number of photons
730
SYMBOLS
p
P
Pa
momentum; probability
power; pressure
pascal (SI unit of pressure)
disintegration energy; electric charge; perfusion; reaction energy
r
R
R
RE
R0
correlation coefficient; radius

roentgen (unit of exposure)
counting rate; dose rate; radius; random coincidence count rate
fractional energy resolution
nuclear radius constant
s
s
scol
srad
stot
s 2
S
Sv
second (unit of time)

septal thickness; sample standard deviation
collision stopping power
radiation stopping power
total stopping power
sample variance
absorbed dose rate per unit activity; sensitivity
sievert (unit of equivalent dose and unit of effective dose)
t time
tesla (SI unit of magnetic field strength)
T
T temperature
T1/2 half-life
u
atomic mass unit
V
V
volt (unit of voltage)

ventilation; voltage; volume
w
wR
wT
W
W
x1/2
x1/10
width
radiation weighting factor
tissue weighting factor
watt (SI unit of power)
mean energy to produce an information carrier; weight
half-value layer
tenth-value layer
x
xe
mean free path

experimental mean
731
SYMBOLS
xt
true mean
X exposure
z
Z
Zeff
specific energy
atomic number
effective atomic number
Greek symbols

alpha particle
linear radiosensitivity constant
beta particle
quadratic radiosensitivity coefficient
gamma ray
air kerma rate constant
0
T

electric constant (permittivity of vacuum)

total energy imparted by radiation to the tissue or organ
mean energy imparted
quantum efficiency; refractive index
pair production attenuation coefficient

pair production atomic attenuation coefficient
radioactive decay constant; wavelength

ab
m
tr
a
e
0
linear attenuation coefficient; mobility of charged carriers

energy absorption coefficient
mass attenuation coefficient
energy transfer coefficient
atomic attenuation coefficient
electronic attenuation coefficient
magnetic constant (permeability of vacuum)
photon frequency
electronic neutrino
electronic antineutrino
e
e
732
SYMBOLS
mass density
cross-section; standard deviation

C
Compton attenuation coefficient
e
experimental standard deviation
eF
experimental fractional standard deviation
F
fractional standard deviation
P
percentage standard deviation
R
Rayleigh attenuation coefficient
Rayleigh atomic attenuation coefficient
aR
Compton atomic attenuation coefficient
aC
Compton electronic attenuation coefficient
eC
2 variance
e2
experimental variance

a
dead time; mean life

photoelectric atomic attenuation coefficient
velocity

angular frequency; fluorescence yield

ohm (SI unit of electrical resistance); solid angle
733
CONTRIBUTORS TO DRAFTING AND REVIEW

Al-Mazrou, R.

Saudi Arabia
Bailey, D.L.
Royal North Shore Hospital and University of Sydney,

Australia
Bergmann, H.
Medical University of Vienna, Austria
Busemann Sokole, E.
Academic Medical Center, Netherlands
Carlsson, S.T.
Uddevalla Hospital, Sweden
Dale, R.G.
Imperial College London, United Kingdom
Daube-Witherspoon, M.E.
University of Pennsylvania, United States of America
Dauer, L.T.

Demirkaya, O.

Saudi Arabia
Du, Yong
Royal Marsden Hospital and Institute of Cancer

Research, United Kingdom
El Fakhri, G.
Massachusetts General Hospital and Harvard Medical

School, United States of America
Flux, G.
Royal Marsden Hospital and Institute of Cancer

Forwood, N.J.
Royal North Shore Hospital, Australia
Frey, E.C.
Johns Hopkins University, United States of America
Hindorf, C.
Skne University Hospital, Sweden
Humm, J.L.

Kesner, A.L.
Le Heron, J.C.
Lodge, M.A.
Johns Hopkins University, United States of America
735
CONTRIBUTORS to drafting and Review
Ltter, M.G. University of the Free State, South Africa

Lundqvist, H.O. Uppsala University, Sweden
Matej, S. University of Pennsylvania, United States of America
Myers, M.J. Imperial College London, United Kingdom
Nuyts, J. Katholieke Universiteit Leuven, Belgium
Ott, R.J. Royal Marsden Hospital and Institute of Cancer
Ouyang, J.
Massachusetts General Hospital and Harvard Medical

School, United States of America
Palm, S. International Atomic Energy Agency

Parker, J.A. Harvard Medical School, United States of America
Podgorsak, E.B.
McGill University, Canada
Poli, G.L. International Atomic Energy Agency

Smart, R.C. St. George Hospital, Australia
Soni, P.S. Bhabha Atomic Research Centre, India
Stephenson, R. Rutherford Appleton Laboratory, United Kingdom
Todd-Pokropek, A. University College London, United Kingdom
van Aswegen, A. University of the Free State, South Africa
Van Eijk, C.W.E. Delft University of Technology, Netherlands
Willowson, K.P. University of Sydney, Australia
Wondergem, J. International Atomic Energy Agency
Zanzonico, P.B.

Consultants Meetings
Vienna, Austria: 15 September 2008, 1416 April 2009, 1719 May 2010, 711 March 2011
736
No. 23
ORDERING LOCALLY
In the following countries, IAEA priced publications may be purchased from the sources listed below or
from major local booksellers.
Orders for unpriced publications should be made directly to the IAEA. The contact details are given at
the end of this list.
AUSTRALIA
DA Information Services
648 Whitehorse Road, Mitcham, VIC 3132, AUSTRALIA
Telephone: +61 3 9210 7777 Fax: +61 3 9210 7788
Email: books@dadirect.com.au Web site: http://www.dadirect.com.au
BELGIUM
Jean de Lannoy
Avenue du Roi 202, 1190 Brussels, BELGIUM
Telephone: +32 2 5384 308 Fax: +32 2 5380 841
Email: jean.de.lannoy@euronet.be Web site: http://www.jean-de-lannoy.be
CANADA
Renouf Publishing Co. Ltd.

5369 Canotek Road, Ottawa, ON K1J 9J3, CANADA
Telephone: +1 613 745 2665 Fax: +1 643 745 7660
Email: order@renoufbooks.com Web site: http://www.renoufbooks.com
Bernan Associates
4501 Forbes Blvd., Suite 200, Lanham, MD 20706-4391, USA
Telephone: +1 800 865 3457 Fax: +1 800 865 3450
Email: orders@bernan.com Web site: http://www.bernan.com
CZECH REPUBLIC
Suweco CZ, spol. S.r.o.

Klecakova 347, 180 21 Prague 9, CZECH REPUBLIC
Telephone: +420 242 459 202 Fax: +420 242 459 203
Email: nakup@suweco.cz Web site: http://www.suweco.cz
FINLAND
Akateeminen Kirjakauppa
PO Box 128 (Keskuskatu 1), 00101 Helsinki, FINLAND
Telephone: +358 9 121 41 Fax: +358 9 121 4450
Email: akatilaus@akateeminen.com Web site: http://www.akateeminen.com
FRANCE
Form-Edit
5 rue Janssen, PO Box 25, 75921 Paris CEDEX, FRANCE
Telephone: +33 1 42 01 49 49 Fax: +33 1 42 01 90 90
Email: fabien.boucard@formedit.fr Web site: http://www.formedit.fr
Lavoisier SAS
14 rue de Provigny, 94236 Cachan CEDEX, FRANCE
Telephone: +33 1 47 40 67 00 Fax: +33 1 47 40 67 02
Email: livres@lavoisier.fr Web site: http://www.lavoisier.fr
LAppel du livre
99 rue de Charonne, 75011 Paris, FRANCE
Telephone: +33 1 43 07 50 80 Fax: +33 1 43 07 50 80
Email: livres@appeldulivre.fr Web site: http://www.appeldulivre.fr
GERMANY
Goethe Buchhandlung Teubig GmbH

Schweitzer Fachinformationen
Willsttterstrasse 15, 40549 Dsseldorf, GERMANY
Telephone: +49 (0) 211 49 8740 Fax: +49 (0) 211 49 87428
Email: s.dehaan@schweitzer-online.de Web site: http://www.goethebuch.de
HUNGARY
Librotade Ltd., Book Import

PF 126, 1656 Budapest, HUNGARY
Telephone: +36 1 257 7777 Fax: +36 1 257 7472
Email: books@librotade.hu Web site: http://www.librotade.hu
INDIA
Allied Publishers
1st Floor, Dubash House, 15, J.N. Heredi Marg, Ballard Estate, Mumbai 400001, INDIA
Telephone: +91 22 2261 7926/27 Fax: +91 22 2261 7928
Email: alliedpl@vsnl.com Web site: http://www.alliedpublishers.com
Bookwell
3/79 Nirankari, Delhi 110009, INDIA
Telephone: +91 11 2760 1283/4536
Email: bkwell@nde.vsnl.net.in Web site: http://www.bookwellindia.com
ITALY
Libreria Scientifica AEIOU

Via Vincenzo Maria Coronelli 6, 20146 Milan, ITALY
Telephone: +39 02 48 95 45 52 Fax: +39 02 48 95 45 48
Email: info@libreriaaeiou.eu Web site: http://www.libreriaaeiou.eu
JAPAN
Maruzen Co., Ltd.

1-9-18 Kaigan, Minato-ku, Tokyo 105-0022, JAPAN
Telephone: +81 3 6367 6047 Fax: +81 3 6367 6160
Email: journal@maruzen.co.jp Web site: http://maruzen.co.jp
NETHERLANDS
Martinus Nijhoff International

Koraalrood 50, Postbus 1853, 2700 CZ Zoetermeer, NETHERLANDS
Telephone: +31 793 684 400 Fax: +31 793 615 698
Email: info@nijhoff.nl Web site: http://www.nijhoff.nl
Swets Information Services Ltd.
PO Box 26, 2300 AA Leiden
Dellaertweg 9b, 2316 WZ Leiden, NETHERLANDS
Telephone: +31 88 4679 387 Fax: +31 88 4679 388
Email: tbeysens@nl.swets.com Web site: http://www.swets.com
SLOVENIA
Cankarjeva Zalozba dd
Kopitarjeva 2, 1515 Ljubljana, SLOVENIA
Telephone: +386 1 432 31 44 Fax: +386 1 230 14 35
Email: import.books@cankarjeva-z.si Web site: http://www.mladinska.com/cankarjeva_zalozba
SPAIN
Diaz de Santos, S.A.

Librerias Bookshop Departamento de pedidos
Calle Albasanz 2, esquina Hermanos Garcia Noblejas 21, 28037 Madrid, SPAIN
Telephone: +34 917 43 48 90 Fax: +34 917 43 4023
Email: compras@diazdesantos.es Web site: http://www.diazdesantos.es
UNITED KINGDOM
The Stationery Office Ltd. (TSO)

PO Box 29, Norwich, Norfolk, NR3 1PD, UNITED KINGDOM
Telephone: +44 870 600 5552
Email (orders): books.orders@tso.co.uk (enquiries): book.enquiries@tso.co.uk Web site: http://www.tso.co.uk
UNITED STATES OF AMERICA
Bernan Associates
4501 Forbes Blvd., Suite 200, Lanham, MD 20706-4391, USA
Telephone: +1 800 865 3457 Fax: +1 800 865 3450
Email: orders@bernan.com Web site: http://www.bernan.com
Renouf Publishing Co. Ltd.
812 Proctor Avenue, Ogdensburg, NY 13669, USA
Telephone: +1 888 551 7470 Fax: +1 888 551 7471
Email: orders@renoufbooks.com Web site: http://www.renoufbooks.com
United Nations
300 East 42nd Street, IN-919J, New York, NY 1001, USA

Telephone: +1 212 963 8302 Fax: 1 212 963 3489
Email: publications@un.org Web site: http://www.unp.un.org
Orders for both priced and unpriced publications may be addressed directly to:
IAEA Publishing Section, Marketing and Sales Unit, International Atomic Energy Agency
Vienna International Centre, PO Box 100, 1400 Vienna, Austria
Telephone: +43 1 2600 22529 or 22488 Fax: +43 1 2600 29302
Email: sales.publications@iaea.org Web site: http://www.iaea.org/books
13-21541
This handbook provides a comprehensive overview of the medical physics

knowledge required in the field of nuclear medicine. It is intended for teachers,
students and residents involved in medical physics programmes. It will serve
as a resource for interested readers from other disciplines, for example,
nuclear medicine physicians, radiochemists and medical technologists, who
would like to familiarize themselves with the basic concepts and practice of
nuclear medicine physics. Physics is a vital aspect of nearly every area of
nuclear medicine, including imaging instrumentation, image processing and
reconstruction, data analysis, radionuclide production, radionuclide therapy,
radiopharmacy, radiation protection and biology. The 20 chapters of this
handbook include a broad coverage of topics relevant to nuclear medicine
physics. The authors and reviewers were drawn from a variety of regions and
were selected because of their knowledge, teaching experience and scientific
acumen. This book was written to address an urgent need for a comprehensive,
contemporary text on the physics of nuclear medicine and has been endorsed
by several international and national organizations. It complements similar
texts in radiation oncology physics and diagnostic radiology physics published
by the IAEA.

Vienna
ISBN 9789201438102

Prirucnik PDF

Transféré par

Informations du document

Titre original

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

Prirucnik PDF

Transféré par

Droits d'auteur :

Formats disponibles

Nuclear Medicine Physics

A Handbook for Teachers and Students

Nuclear Medicine Physics:

Nuclear Medicine Physics:

International atomic energy agency

IAEA Library Cataloguing in Publication Data

PHOTON INTERACTIONS WITH MATTER. . . . . . . . . . . . . . 30

CHAPTER 2. BASIC RADIOBIOLOGY. . . . . . . . . . . . . . . . . . . . . . . . . . . 49

GROSS RADIATION EFFECTS ON TUMOURS AND

CHAPTER 3. RADIATION PROTECTION. . . . . . . . . . . . . . . . . . . . . . . . . 73

CHAPTER 4. RADIONUCLIDE PRODUCTION. . . . . . . . . . . . . . . . . . . . 117

THE ORIGINS OF DIFFERENT NUCLEI. . . . . . . . . . . . . . . . . 117

4.2.2. Thermal and fast neutron reactions . . . . . . . . . . . . . . . . 128

CHAPTER 5. STATISTICS FOR RADIATION MEASUREMENT . . . . . . 149

SOURCES OF ERROR IN NUCLEAR MEDICINE

PROPAGATION OF ERROR . . . . . . . . . . . . . . . . . . . . . . . . . . . 172

CHAPTER 6. BASIC RADIATION DETECTORS . . . . . . . . . . . . . . . . . . . 196

CHAPTER 7. ELECTRONICS RELATED TO NUCLEAR

CHAPTER 8. GENERIC PERFORMANCE MEASURES . . . . . . . . . . . . . 234

INTRINSIC AND EXTRINSIC MEASURES. . . . . . . . . . . . . . . 234

SPATIAL RESOLUTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240

CHAPTER 9. PHYSICS IN THE RADIOPHARMACY. . . . . . . . . . . . . . . 251

THE MODERN RADIONUCLIDE CALIBRATOR. . . . . . . . . . 251

PRODUCT CONTAINMENT ENCLOSURES. . . . . . . . . . . . . . 271

CHAPTER 10. NON-IMAGING DETECTORS AND COUNTERS. . . . . . . 287

CHAPTER 11. NUCLEAR MEDICINE IMAGING DEVICES. . . . . . . . . . 312

12.4.3. SPECT. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 427

CHAPTER 14. NUCLEAR MEDICINE IMAGE DISPLAY. . . . . . . . . . . . . 512

15.4. ACCEPTANCE TESTING. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 578

18.1.3. Absorbed dose rate per unit activity (S value). . . . . . . . 628

20.5. RADIONUCLIDE THERAPY TREATMENT ROOMS

Length l: metre (m).

BASIC PHYSICS FOR NUCLEAR MEDICINE

(e) Temperature T: kelvin (K).

1 m=109 nm=1010 =1015 fm

1 s=103 ms=106 s=109 ns=1012 ps

1 A=103 mA=106 A=109 nA

T (in K)=T (in C) + 273.16

760 torr=101.3 kPa 1 Pa=1 N/m2 =7.5 103 torr

1 eV=1.602 1019 J =103 keV

1.1.3. Classification of radiation

BASIC PHYSICS FOR NUCLEAR MEDICINE

1.1.5. Classification of indirectly ionizing photon radiation

a central role in modern imaging and therapeutic equipment, since it can be

BASIC PHYSICS FOR NUCLEAR MEDICINE

1.1.10. Radiation quantities and units

1.2. BASIC DEFINITIONS FOR ATOMIC STRUCTURE

BASIC PHYSICS FOR NUCLEAR MEDICINE

The number of electrons per volume of an element is:

The number of electrons per mass of an element is:

1.2.1. Rutherford model of the atom

BASIC PHYSICS FOR NUCLEAR MEDICINE

Since then, fission reactors have become an important means of production

BASIC PHYSICS FOR NUCLEAR MEDICINE

The Q value of a nuclear reaction is defined as the difference between the

FIG. 1.2. Schematic representation of a two-particle collision of a projectile (incident

BASIC PHYSICS FOR NUCLEAR MEDICINE

Specific activity a is defined as activity A per unit mass m, i.e.:

where NA is Avogadros number.

The rate of depletion of the number of radioactive parent nuclei NP(t) is

The fundamental differential equation in Eq.(1.10) for NP(t) can be

where NP(0) is the initial condition represented by the number of radioactive

Assuming that P is constant, Eq.(1.11) can be solved to obtain:

and Z n is the number of electrons per unit volume of absorber,